📖 Tier 1: Prepare & Study Guide
✓ 100% Solved with Rationales
Artificial Intelligence Solved Questions Bank & Study Material (2026) - Apex Rankers
Artificial Intelligence & Data Science | Topic-Wise Comprehensive Solved Pool
410 Total Solved Questions
~615 mins Estimated Reading Time
5 Subject Areas / Chapters
Select Topic Area / Chapter:
Click any section below to switch questions
Classical Machine Learning & Statistical Foundations
Showing 25 of 82 (30%)
Jump:
🎯 Practice This Area
🎉 All 82 Questions Loaded!
Computer Vision & Geospatial Remote Sensing AI
Showing 25 of 82 (30%)
Jump:
🎯 Practice This Area
🎉 All 82 Questions Loaded!
Deep Learning Architectures & Optimization
Showing 25 of 82 (30%)
Jump:
🎯 Practice This Area
🎉 All 82 Questions Loaded!
Time-Series Forecasting & MLOps Deployment
Showing 25 of 82 (30%)
Jump:
🎯 Practice This Area
🎉 All 82 Questions Loaded!
Transformers, LLMs & Natural Language Processing
Showing 25 of 82 (30%)
Jump:
🎯 Practice This Area
🎉 All 82 Questions Loaded!
Q. 1
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In linear regression, what fundamental assumption is violated when the variance of the residuals is not constant across all levels of the independent variables?
A
Homoscedasticity (resulting in Heteroscedasticity)
✓ Correct
B
Multicollinearity
C
Normality of features
D
Linear independence of predictors
💡
Step-by-Step Explanation & Concept Rationale
Heteroscedasticity occurs when error terms have non-constant variance, leading to inefficient OLS coefficient standard error estimates.
Q. 2
Artificial Intelligence
Difficulty: Medium
(1 Mark)
Which regularization technique adds a penalty equal to the sum of the absolute values of the coefficients (L1 penalty) to the loss function, inducing feature sparsity?
A
Lasso Regression (L1 Regularization)
✓ Correct
B
Ridge Regression (L2 Regularization)
C
Ordinary Least Squares
D
Polynomial Regression
💡
Step-by-Step Explanation & Concept Rationale
Lasso adds lambda * sum(|beta_j|), driving irrelevant feature weights to exactly zero, thus performing automated feature selection.
Q. 3
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the primary difference between L1 (Lasso) and L2 (Ridge) regularization?
A
L1 penalizes absolute magnitude and can set weights to zero; L2 penalizes squared magnitude and shrinks weights toward zero without setting them strictly to zero
✓ Correct
B
L1 is for classification only; L2 is for regression only
C
L1 increases model variance; L2 increases model bias
D
L1 is non-differentiable everywhere; L2 is non-convex
💡
Step-by-Step Explanation & Concept Rationale
L1 uses diamond-shaped contour constraints promoting sparse solutions; L2 uses spherical contours shrinking weights smoothly.
Q. 4
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What does the 'Bias-Variance Tradeoff' represent in machine learning?
A
The conflict between an algorithm's error from erroneous assumptions (Bias) and its sensitivity to small fluctuations in training data (Variance)
✓ Correct
B
The trade-off between training speed and inference latency
C
The balance between dataset size and RAM usage
D
The compromise between CPU and GPU hardware costs
💡
Step-by-Step Explanation & Concept Rationale
High bias causes underfitting (oversimplified model); high variance causes overfitting (capturing noise). Total error = Bias^2 + Variance + Irreducible Noise.
Q. 5
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In classification problems, which metric is defined as the harmonic mean of Precision and Recall?
A
F1-Score
✓ Correct
B
ROC-AUC
C
Accuracy
D
Cohen's Kappa
💡
Step-by-Step Explanation & Concept Rationale
F1 = 2 * (Precision * Recall) / (Precision + Recall). It balances false positives and false negatives, especially in imbalanced datasets.
Q. 6
Artificial Intelligence
Difficulty: Medium
(1 Mark)
When evaluating a model on a highly imbalanced dataset (e.g. 99% non-flood, 1% flood disaster events), why is Accuracy a misleading performance metric?
A
A naive model predicting 'no flood' for all instances achieves 99% accuracy while completely failing to detect any real flood event
✓ Correct
B
Accuracy cannot be computed for binary classification
C
Accuracy only works for regression tasks
D
Accuracy requires all features to be standardized
💡
Step-by-Step Explanation & Concept Rationale
Accuracy paradox: high accuracy masks severe failure on the minority class. PR-AUC, F1-score, or Recall must be prioritized.
Q. 7
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Precision' in a binary classification confusion matrix?
A
True Positives / (True Positives + False Positives)
✓ Correct
B
True Positives / (True Positives + False Negatives)
C
True Negatives / (True Negatives + False Positives)
D
True Positives + True Negatives / Total
💡
Step-by-Step Explanation & Concept Rationale
Precision (Positive Predictive Value) measures the proportion of predicted positive instances that were actually true positives.
Q. 8
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Recall' (also known as Sensitivity or True Positive Rate)?
A
True Positives / (True Positives + False Negatives)
✓ Correct
B
True Positives / (True Positives + False Positives)
C
False Positives / (False Positives + True Negatives)
D
True Negatives / (True Negatives + False Negatives)
💡
Step-by-Step Explanation & Concept Rationale
Recall measures the model's ability to find all actual positive instances in the dataset.
Q. 9
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What does an Area Under the ROC Curve (ROC-AUC) score of 0.5 indicate?
A
The classifier has performance equivalent to random guessing
✓ Correct
B
The classifier has perfect discrimination ability
C
The classifier makes 100% false positive predictions
D
The model is severely overfitted
💡
Step-by-Step Explanation & Concept Rationale
ROC-AUC plots TPR vs FPR across all classification thresholds; 0.5 is random chance, 1.0 is perfect classification.
Q. 10
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In decision trees, which impurity measure is computed as 1 - sum(p_i^2) for class probabilities p_i?
A
Gini Impurity
✓ Correct
B
Shannon Entropy
C
Information Gain Ratio
D
Mean Squared Error
💡
Step-by-Step Explanation & Concept Rationale
Gini impurity measures the probability of misclassifying a randomly chosen element; it is the default criterion in CART trees.
Q. 11
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the core ensemble mechanism behind 'Random Forest'?
A
Bagging (Bootstrap Aggregating) combined with random feature subspace selection at each split
✓ Correct
B
Sequential residual boosting with gradient descent
C
Linear stacking with meta-classifiers
D
Iterative hard voting on single decision trees
💡
Step-by-Step Explanation & Concept Rationale
Random Forest trains diverse decorrelated decision trees on bootstrap samples and aggregates their predictions via majority voting or averaging.
Q. 12
Artificial Intelligence
Difficulty: Hard
(1 Mark)
How does 'Gradient Boosting' (e.g. GBM, XGBoost, LightGBM) build its ensemble differently from Random Forest?
A
It trains trees sequentially, where each new tree is fitted to predict the pseudo-residuals (negative gradient of the loss function) of the previous ensemble
✓ Correct
B
It trains all trees completely in parallel on random bootstrap subsets
C
It drops random nodes during training like Dropout
D
It clusters training data using k-Means before fitting trees
💡
Step-by-Step Explanation & Concept Rationale
Boosting is additive and sequential; each base learner compensates for the errors of the preceding ensemble.
Q. 13
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What key innovation allows XGBoost to achieve superior speed and accuracy compared to standard gradient boosting?
A
Second-order Taylor expansion of the loss function (Hessian), exact/approximate greedy split finding, built-in L1/L2 regularization, and sparsity-aware split finding
✓ Correct
B
Using neural network backpropagation exclusively
C
Eliminating the need for decision trees
D
Running only on quantum computers
💡
Step-by-Step Explanation & Concept Rationale
XGBoost incorporates first (gradient) and second (Hessian) derivatives in objective optimization along with shrinkage and column subsampling.
Q. 14
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the primary architectural feature of LightGBM that makes it faster on large datasets than traditional gradient boosted trees?
A
Leaf-wise (best-first) tree growth, Histogram-based binning, Gradient-based One-Side Sampling (GOSS), and Exclusive Feature Bundling (EFB)
✓ Correct
B
Level-wise depth-first tree growth exclusively
C
Using radial basis functions in leaf nodes
D
Converting all numerical features into raw text strings
💡
Step-by-Step Explanation & Concept Rationale
LightGBM grows trees leaf-wise rather than level-wise, achieving lower loss with GOSS and histogram binning reducing memory and compute.
Q. 15
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In Support Vector Machines (SVM), what is the 'Kernel Trick'?
A
Implicitly mapping input data into a higher-dimensional feature space where it becomes linearly separable, using inner product kernel functions without calculating coordinates explicitly
✓ Correct
B
Removing outliers by trimming extreme data points
C
Converting continuous labels into categorical classes
D
Optimizing weights using stochastic gradient descent
💡
Step-by-Step Explanation & Concept Rationale
The kernel trick computes inner products in high-dimensional Hilbert spaces using functions like RBF/Gaussian, Polynomial, or Sigmoid.
Q. 16
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In SVM, what role does the hyperparameter 'C' play in the soft-margin formulation?
A
It controls the trade-off between maximizing the margin width and minimizing classification margin violations (slack errors)
✓ Correct
B
It sets the learning rate for gradient descent
C
It defines the number of decision trees in the ensemble
D
It sets the batch size during training
💡
Step-by-Step Explanation & Concept Rationale
Large C penalizes margin violations heavily (narrow margin, risk of overfitting); small C allows more margin violations (wider margin, higher bias).
Q. 17
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In k-Nearest Neighbors (k-NN) classification, how does increasing the value of 'k' affect the model's decision boundary?
A
It smoothens the decision boundary, reducing model variance and increasing bias
✓ Correct
B
It makes the decision boundary more jagged and complex
C
It causes severe overfitting on training data
D
It reduces the training time to zero
💡
Step-by-Step Explanation & Concept Rationale
A small k (e.g. k=1) creates complex, noisy boundaries (overfitting); large k averages over broader neighborhoods (smoother, higher bias).
Q. 18
Artificial Intelligence
Difficulty: Medium
(1 Mark)
Why is feature scaling (e.g. StandardScaler or MinMaxScaler) mandatory prior to training distance-based algorithms like k-NN and k-Means?
A
Features with larger numerical ranges will dominate Euclidean distance calculations, rendering features with smaller scales virtually ineffective
✓ Correct
B
Distance algorithms will fail to compile without scaling
C
Scaling converts categorical features to integers
D
Scaling eliminates all missing values in data
💡
Step-by-Step Explanation & Concept Rationale
Euclidean distance is sensitive to scale; a feature in thousands (e.g. income) dwarfs a feature in decimals (e.g. ratios) unless standardized.
Q. 19
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the fundamental assumption of the 'Naive Bayes' classifier that gives it the name 'Naive'?
A
All features are mutually independent of each other given the class label
✓ Correct
B
All features follow a perfectly uniform distribution
C
The target classes are always perfectly balanced
D
The data contains no outliers or missing values
💡
Step-by-Step Explanation & Concept Rationale
Naive Bayes assumes conditional independence: P(x1, x2 | y) = P(x1 | y) * P(x2 | y), which simplifies probability computation drastically.
Q. 20
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In k-Means clustering, what does the 'Elbow Method' evaluate to help select the optimal number of clusters (k)?
A
The Within-Cluster Sum of Squares (WCSS / Inertia) plotted against k, looking for the point of diminishing returns
✓ Correct
B
The classification accuracy on holdout test set
C
The F1-score across multiple clusters
D
The gradient norm during backpropagation
💡
Step-by-Step Explanation & Concept Rationale
As k increases, WCSS decreases; the 'elbow' inflection point identifies where adding more clusters yields marginal variance reduction.
Q. 21
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What metric measures how similar an object is to its own cluster compared to other clusters, ranging from -1 to +1?
A
Silhouette Coefficient (Silhouette Score)
✓ Correct
B
Adjusted Rand Index
C
Davies-Bouldin Index
D
Calinski-Harabasz Index
💡
Step-by-Step Explanation & Concept Rationale
Silhouette score: (b - a) / max(a, b), where a is mean intra-cluster distance and b is mean nearest-cluster distance. +1 indicates well-clustered samples.
Q. 22
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the primary advantage of DBSCAN (Density-Based Spatial Clustering of Applications with Noise) over k-Means?
A
It can discover clusters of arbitrary shapes and automatically identifies noise/outliers without requiring the number of clusters to be pre-specified
✓ Correct
B
It guarantees spherical clusters of identical diameter
C
It operates in linear O(1) time complexity
D
It works exclusively on text data
💡
Step-by-Step Explanation & Concept Rationale
DBSCAN groups points based on density reachability (eps and min_samples), isolating sparse noise points as outliers.
Q. 23
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In Principal Component Analysis (PCA), how are the principal components constructed from the data covariance matrix?
A
As the eigenvectors corresponding to the largest eigenvalues of the covariance matrix
✓ Correct
B
As random orthogonal projections across features
C
As non-linear kernel transformations minimizing cross-entropy
D
As decision tree leaf node assignments
💡
Step-by-Step Explanation & Concept Rationale
Eigenvectors define the principal axes of maximum variance, while eigenvalues quantify the variance explained along each component.
Q. 24
Artificial Intelligence
Difficulty: Medium
(1 Mark)
Why must features be centered (zero mean) and scaled before performing Principal Component Analysis (PCA)?
A
To ensure that high-variance unscaled features do not artificially dominate the first principal component
✓ Correct
B
PCA cannot compute matrix determinants without scaling
C
To convert categorical features into continuous variables
D
To ensure the data follows a Poisson distribution
💡
Step-by-Step Explanation & Concept Rationale
PCA maximizes variance; features with larger raw magnitudes will otherwise dominate the covariance matrix regardless of information content.
Q. 25
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the primary difference between PCA and t-SNE (t-Distributed Stochastic Neighbor Embedding)?
A
PCA is a linear technique preserving global variance; t-SNE is a non-linear probabilistic technique preserving local neighborhood structures (ideal for 2D/3D visualization)
✓ Correct
B
PCA is supervised; t-SNE is unsupervised
C
PCA is for text only; t-SNE is for images only
D
PCA requires GPUs; t-SNE runs on microcontrollers
💡
Step-by-Step Explanation & Concept Rationale
t-SNE minimizes the Kullback-Leibler divergence between high-dimensional joint probabilities and low-dimensional Student-t probabilities.
Study Stream Progress:
Showing 25 of 82 Questions (30%)
Jump to:
🎉
All 82 Solved Questions in Classical Machine Learning & Statistical Foundations Loaded!
You have studied the complete question pool with step-by-step rationales. Ready to test yourself?
Q. 1
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the primary difference between Two-Stage and One-Stage Object Detectors?
A
Two-stage detectors (Faster R-CNN) first generate Region Proposals (RPN) and then classify/refine them; One-stage detectors (YOLO, SSD) predict bounding boxes and class probabilities directly in a single dense grid pass
✓ Correct
B
Two-stage detectors process images in black and white; One-stage detectors in color
C
Two-stage detectors run on CPU; One-stage detectors on GPU
D
Two-stage detectors only detect 2 objects per image
💡
Step-by-Step Explanation & Concept Rationale
Two-stage models prioritize detection accuracy/mAP; one-stage models achieve real-time inference speeds (>60 FPS) suitable for video streams and drone feeds.
Q. 2
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In 'Faster R-CNN', what is the role of the 'Region Proposal Network' (RPN)?
A
A lightweight convolutional network that slides over feature maps to generate candidate bounding box proposals (RoIs) with objectness scores and anchor box offsets
✓ Correct
B
A module that scales image brightness
C
A module that converts bounding boxes to text captions
D
A classifier that performs OCR on road signs
💡
Step-by-Step Explanation & Concept Rationale
RPN replaced slow CPU selective search with fast GPU-based anchor proposals, sharing convolutional features with the downstream detection network.
Q. 3
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In YOLO (You Only Look Once), what is the 'Non-Maximum Suppression' (NMS) post-processing algorithm?
A
An iterative algorithm that filters out overlapping redundant bounding boxes, keeping only the highest confidence box and eliminating boxes with IoU (Intersection over Union) > threshold
✓ Correct
B
A method to suppress dark background pixels in images
C
A neural network layer that speeds up backpropagation
D
A technique for cropping images to square shapes
💡
Step-by-Step Explanation & Concept Rationale
NMS cleans up dense multi-box detections, ensuring each unique physical object is localized with exactly one bounding box.
Q. 4
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Intersection over Union' (IoU) / Jaccard Index in object detection and semantic segmentation evaluation?
A
Area of Overlap / Area of Union: |A intersect B| / |A union B| between predicted bounding box/mask and ground-truth annotation
✓ Correct
B
The sum of image width and height
C
The ratio of precision to recall
D
The angle between two bounding box diagonals
💡
Step-by-Step Explanation & Concept Rationale
IoU >= 0.5 is the standard threshold for True Positive detection in PASCAL VOC and COCO mAP evaluations.
Q. 5
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Mean Average Precision' (mAP, e.g. mAP@0.5, mAP@0.5:0.95) in object detection?
A
The mean of the Average Precision (area under the Precision-Recall curve) calculated across all object categories and IoU threshold intervals
✓ Correct
B
The average classification accuracy across all pixels
C
The speed of object detection in frames per second
D
The mean pixel intensity across RGB channels
💡
Step-by-Step Explanation & Concept Rationale
COCO mAP@0.5:0.95 averages AP across 10 IoU thresholds (from 0.50 to 0.95 with step 0.05) to rigorously penalize localization inaccuracies.
Q. 6
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'U-Net' architecture (Ronneberger et al. 2015) and why is it the gold standard for biomedical and satellite image segmentation?
A
An encoder-decoder network featuring skip connections that concatenate high-resolution spatial feature maps from contracting paths directly to expanding upsampling paths
✓ Correct
B
A network shaped like a circle that loops infinitely
C
A generative network that produces synthetic images from text
D
A network that processes 1D audio waves
💡
Step-by-Step Explanation & Concept Rationale
U-Net skip connections preserve precise pixel-level spatial boundaries lost during downsampling, essential for flood inundation and tumor boundary segmentation.
Q. 7
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the primary difference between 'Semantic Segmentation' and 'Instance Segmentation'?
A
Semantic segmentation labels every pixel with a class (e.g. all people are red); Instance segmentation labels every pixel AND differentiates distinct object instances (e.g. Person 1 is red, Person 2 is blue)
✓ Correct
B
Semantic is for text; Instance is for images
C
Semantic is 2D; Instance is 3D
D
There is no difference between them
💡
Step-by-Step Explanation & Concept Rationale
Instance segmentation (e.g. Mask R-CNN, SAM) detects individual objects with distinct mask boundaries, whereas semantic segmentation (e.g. DeepLab) groups all instances into one class mask.
Q. 8
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In 'Mask R-CNN', what operation replaced 'RoIPool' to eliminate spatial quantization errors and achieve sub-pixel mask alignment?
A
RoIAlign (using bilinear interpolation to extract feature values at continuous floating-point locations without discrete coordinate rounding)
✓ Correct
B
Max Pooling with stride 4
C
Global Average Pooling
D
Batch Normalization
💡
Step-by-Step Explanation & Concept Rationale
He et al. showed RoIPool's coordinate rounding introduced substantial spatial misalignments; RoIAlign resolves this with continuous bilinear sampling.
Q. 9
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Vision Transformer' (ViT, Dosovitskiy et al. 2020)?
A
An architecture that splits an image into a sequence of non-overlapping patches (e.g. 16x16 pixels), projects them linearly into token embeddings, and processes them with standard Transformer Encoders
✓ Correct
B
A transformer used in optical fiber internet cables
C
A camera lens that zooms automatically using AI
D
A convolutional network that processes text prompts
💡
Step-by-Step Explanation & Concept Rationale
ViT showed that standard Transformer architectures with minimal inductive bias surpass CNNs on large pre-training scales (JFT-300M, ImageNet-21k).
Q. 10
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Swin Transformer' (Hierarchical Vision Transformer using Shifted Windows)?
A
A vision transformer that computes self-attention within local non-overlapping windows and shifts window partitioning across layers to achieve linear O(N) computational complexity
✓ Correct
B
A transformer that swings back and forth during training
C
A vision model trained on swimming pool images
D
A model that processes images in reverse order
💡
Step-by-Step Explanation & Concept Rationale
Swin Transformer provides multi-scale hierarchical feature maps and linear computational complexity relative to image size, making it a powerful general backbone for dense prediction.
Q. 11
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Segment Anything Model' (SAM, Kirillov et al. / Meta 2023)?
A
A foundation model for zero-shot promptable image segmentation that generates high-quality masks from points, bounding boxes, or free-form text prompts
✓ Correct
B
A computer software that edits video audio tracks
C
A tool for segmenting hard drive disk partitions
D
A database partitioning tool for PostgreSQL
💡
Step-by-Step Explanation & Concept Rationale
SAM was trained on 1 billion masks (SA-1B), enabling zero-shot generalization to segment any object in natural, medical, or satellite imagery.
Q. 12
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In multispectral satellite remote sensing (e.g. Sentinel-2, Landsat-8), what is the 'Normalized Difference Vegetation Index' (NDVI) formula?
A
NDVI = (NIR - Red) / (NIR + Red)
✓ Correct
B
NDVI = (Red - Blue) / (Red + Blue)
C
NDVI = (SWIR - Green) / (SWIR + Green)
D
NDVI = (Thermal - Visible) / (Thermal + Visible)
💡
Step-by-Step Explanation & Concept Rationale
Healthy green vegetation absorbs red light (chlorophyll absorption) and strongly reflects Near-Infrared (NIR); NDVI ranges from -1 to +1.
Q. 13
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What index is universally calculated from Near-Infrared (NIR) and Green bands to detect and map standing flood waters and open water bodies from satellite imagery?
A
Normalized Difference Water Index (NDWI = (Green - NIR) / (Green + NIR))
✓ Correct
B
Normalized Difference Vegetation Index (NDVI)
C
Enhanced Vegetation Index (EVI)
D
Soil Adjusted Vegetation Index (SAVI)
💡
Step-by-Step Explanation & Concept Rationale
McFeeters NDWI uses Green and NIR to delineate open water surfaces; water has high reflectance in green and near-total absorption in NIR.
Q. 14
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the 'Normalized Difference Built-up Index' (NDBI) used for in satellite urban analysis?
A
NDBI = (SWIR - NIR) / (SWIR + NIR), used to delineate urban impervious infrastructure, concrete, and built-up settlements
✓ Correct
B
Measuring forest tree density
C
Detecting ocean oil spills
D
Mapping agricultural crop moisture
💡
Step-by-Step Explanation & Concept Rationale
Built-up concrete and urban materials exhibit higher reflectance in Shortwave Infrared (SWIR) compared to NIR.
Q. 15
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Synthetic Aperture Radar' (SAR) (e.g. Sentinel-1 C-band SAR) and why is it indispensable for flood disaster monitoring in cloudy weather?
A
An active microwave sensor that penetrates thick storm clouds, heavy monsoon rain, and operates day or night, providing all-weather surface radar backscatter
✓ Correct
B
An optical digital camera that takes color photos in sunlight
C
A thermal infrared camera measuring land surface heat
D
A sound wave sensor measuring acoustic noise
💡
Step-by-Step Explanation & Concept Rationale
Optical satellites cannot see the ground through cloud cover during active monsoon storms; SAR microwaves penetrate clouds and detect flood water as dark specular reflectors.
Q. 16
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In SAR flood imagery analysis, why do calm standing flood waters appear as dark (low backscatter) pixels in radar images?
A
Smooth standing water acts as a specular reflector, bouncing radar microwave pulses away from the satellite antenna, resulting in near-zero signal return
✓ Correct
B
Water absorbs all radar energy into thermal heat
C
Water reflects radar pulses back directly to the satellite
D
Water freezes into ice under radar pulses
💡
Step-by-Step Explanation & Concept Rationale
Specular reflection directs the radar beam away from the sensor (low dB backscatter), creating sharp contrast against rough, bright land surfaces.
Q. 17
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Change Detection' (e.g. Bi-Temporal Siamese Networks) in satellite disaster damage assessment?
A
Comparing pre-disaster and post-disaster satellite imagery pairs to identify damaged buildings, collapsed bridges, and inundated roads using deep neural feature differences
✓ Correct
B
Detecting changes in computer operating system files
C
Checking weather temperature changes over 24 hours
D
Measuring seasonal crop planting cycles
💡
Step-by-Step Explanation & Concept Rationale
Siamese networks extract deep feature representations from registered pre/post image pairs to classify structural collapse and flood inundation.
Q. 18
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Spatial Resolution' in satellite remote sensing (e.g. Sentinel-2 at 10m vs WorldView-3 at 0.3m)?
A
The physical ground area represented by a single image pixel (e.g. 10m means 1 pixel covers a 10m x 10m ground footprint)
✓ Correct
B
The total number of spectral color bands
C
The frequency of satellite orbital flyovers
D
The file size of the downloaded satellite image
💡
Step-by-Step Explanation & Concept Rationale
Higher spatial resolution (smaller pixel size in meters) allows detection of smaller physical ground objects (individual tents vs broad regional floods).
Q. 19
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Temporal Resolution' (Revisit Time) in satellite constellations?
A
The time elapsed between consecutive satellite observations of the exact same geographical point on Earth (e.g. 5 days for Sentinel-2, daily for PlanetScope)
✓ Correct
B
The shutter exposure speed of the satellite camera
C
The lifespan of the satellite battery in orbit
D
The time taken to download data from the cloud
💡
Step-by-Step Explanation & Concept Rationale
High temporal resolution (frequent revisits) is critical for tracking dynamic, rapid-onset flash floods and cyclone paths in real time.
Q. 20
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Spectral Resolution' in earth observation sensors?
A
The number, width, and wavelength range of specific electromagnetic spectral bands (e.g. RGB, RedEdge, NIR, SWIR, Thermal) captured by the sensor
✓ Correct
B
The spatial size of the camera sensor chip
C
The zoom magnification level of the optical lens
D
The frame rate of video recording
💡
Step-by-Step Explanation & Concept Rationale
Hyperspectral sensors have hundreds of narrow contiguous bands; multispectral sensors have 10-15 broad bands.
Q. 21
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Radiometric Resolution' in remote sensing sensors (e.g. 8-bit vs 12-bit vs 16-bit)?
A
The sensor's sensitivity to subtle differences in reflected electromagnetic energy, determined by the number of digital quantization levels (e.g. 12-bit = 4,096 levels)
✓ Correct
B
The physical radio frequency of satellite downlink
C
The radiation shielding thickness of the satellite casing
D
The power wattage of solar panels
💡
Step-by-Step Explanation & Concept Rationale
Higher radiometric resolution allows distinguishing subtle moisture and soil variations in shadow or dark terrain.
Q. 22
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Atmospheric Correction' (e.g. BOA / Level-2A surface reflectance) in optical satellite preprocessing?
A
Removing atmospheric scattering, aerosol absorption, and Rayleigh distortion to convert Top-Of-Atmosphere (TOA) radiance into true Surface Reflectance (BOA)
✓ Correct
B
Warming up the satellite sensor with heaters
C
Cleaning dust off the satellite camera lens
D
Adjusting camera shutter speed for wind gusts
💡
Step-by-Step Explanation & Concept Rationale
Atmospheric correction ensures reflectance values are physically comparable across different dates and seasons.
Q. 23
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Cloud Masking' (e.g. Fmask / SCL in Sentinel-2) prior to geospatial AI analysis?
A
Identifying and filtering out clouds, cirrus, and cloud shadows from satellite imagery so they are not misclassified as floodwater, snow, or urban structures
✓ Correct
B
Drawing artificial clouds onto images for aesthetics
C
Masking cloud computing IP addresses
D
Encrypting satellite data files in the cloud
💡
Step-by-Step Explanation & Concept Rationale
Cloud shadows have low reflectance and can be easily confused with floodwater unless masked out.
Q. 24
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Pansharpening' in satellite image processing?
A
Fusing a high-resolution panchromatic (grayscale) band with lower-resolution multispectral color bands to produce a high-resolution color composite image
✓ Correct
B
Sharpening a physical kitchen frying pan
C
Increasing image brightness uniformly
D
Cropping an image with a rectangular marquee tool
💡
Step-by-Step Explanation & Concept Rationale
Pansharpening blends the high spatial detail of panchromatic sensors with the rich spectral data of multispectral sensors.
Q. 25
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Object-Based Image Analysis' (OBIA) compared to Pixel-Based Classification?
A
Grouping adjacent pixels into homogeneous vector image segments (superpixels) based on texture, shape, and context before classification, reducing 'salt-and-pepper' noise
✓ Correct
B
Classifying pixels completely independently without spatial context
C
Deleting objects from an image
D
Drawing vector bounding boxes manually
💡
Step-by-Step Explanation & Concept Rationale
OBIA uses algorithms like SLIC or multiresolution segmentation to mimic human visual cognitive interpretation.
Study Stream Progress:
Showing 25 of 82 Questions (30%)
Jump to:
🎉
All 82 Solved Questions in Computer Vision & Geospatial Remote Sensing AI Loaded!
You have studied the complete question pool with step-by-step rationales. Ready to test yourself?
Q. 1
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the primary cause of the 'Vanishing Gradient Problem' during backpropagation in deep neural networks when using Sigmoid or Tanh activation functions?
A
The derivatives of Sigmoid and Tanh are strictly less than 1 (maximum 0.25 for Sigmoid), causing error gradients to shrink exponentially as they are multiplied through multiple layers
✓ Correct
B
The learning rate is set too high, causing weights to oscillate to infinity
C
The dataset contains negative numbers only
D
The GPU runs out of video memory during matrix multiplication
💡
Step-by-Step Explanation & Concept Rationale
By the chain rule, multiplying numbers < 1 across many layers causes gradients for early layers to vanish toward zero, halting weight updates.
Q. 2
Artificial Intelligence
Difficulty: Medium
(1 Mark)
How does the Rectified Linear Unit (ReLU: f(x) = max(0, x)) activation function mitigate the vanishing gradient problem in deep networks?
A
For all positive inputs (x > 0), the gradient is constant at 1.0, allowing error signals to flow backward through deep layers without exponential decay
✓ Correct
B
It bounds outputs between -1 and +1
C
It uses exponential functions to amplify gradients
D
It calculates second-order derivatives automatically
💡
Step-by-Step Explanation & Concept Rationale
ReLU's constant gradient of 1 for x > 0 eliminates gradient saturation in the positive regime.
Q. 3
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the 'Dying ReLU' problem and what architectural modification directly addresses it?
A
Neurons whose inputs are consistently negative output 0 with 0 gradient and permanently stop learning; addressed by Leaky ReLU (f(x) = max(alpha * x, x)) or PReLU
✓ Correct
B
Neurons overheating the GPU processor; addressed by liquid cooling
C
Weights becoming NaN due to division by zero; addressed by adding epsilon
D
Loss function diverging to infinity; addressed by weight decay
💡
Step-by-Step Explanation & Concept Rationale
Leaky ReLU assigns a small non-zero slope (e.g. 0.01) for x < 0, allowing gradients to flow and revive dead neurons.
Q. 4
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What activation function, defined as x * Phi(x) where Phi is the standard Gaussian cumulative distribution function, is widely used in modern Transformers like BERT and GPT?
A
Gaussian Error Linear Unit (GELU)
✓ Correct
B
Sigmoid
C
Hard Tanh
D
Binary Step Function
💡
Step-by-Step Explanation & Concept Rationale
GELU weights inputs by their probability under a Gaussian distribution, providing smooth non-linear probabilistic gating.
Q. 5
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the fundamental purpose of 'Backpropagation' in artificial neural networks?
A
Efficiently computing the partial derivatives (gradients) of the loss function with respect to every learnable weight in the network using the calculus Chain Rule
✓ Correct
B
Sorting training data in ascending order
C
Generating random initial weights for the network
D
Converting images into grayscale matrices
💡
Step-by-Step Explanation & Concept Rationale
Backpropagation propagates errors backward from the output layer to compute dLoss/dWeight for gradient descent optimization.
Q. 6
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In stochastic gradient descent, what is the role of the 'Momentum' term?
A
Accelerating gradient descent in the relevant direction and dampening oscillations by accumulating an exponentially decaying moving average of past gradients
✓ Correct
B
Randomly resetting weights to zero every epoch
C
Increasing the batch size dynamically
D
Normalizing input pixel values
💡
Step-by-Step Explanation & Concept Rationale
Momentum v_t = gamma * v_{t-1} + eta * grad helps the optimizer navigate valleys and push past flat local minima or saddle points.
Q. 7
Artificial Intelligence
Difficulty: Hard
(1 Mark)
How does the 'Adam' (Adaptive Moment Estimation) optimizer compute parameter updates?
A
By maintaining exponentially decaying moving averages of both past gradients (first moment / mean) and past squared gradients (second moment / uncentered variance)
✓ Correct
B
By computing exact second-order Hessian matrices at every step
C
By using random walk exploration without gradients
D
By keeping learning rates completely constant for all weights
💡
Step-by-Step Explanation & Concept Rationale
Adam combines the benefits of AdaGrad (handling sparse gradients) and RMSprop (handling non-stationary objectives) with bias-corrected moment estimates.
Q. 8
Artificial Intelligence
Difficulty: Hard
(1 Mark)
Why is 'AdamW' (Adam with Decoupled Weight Decay) preferred over standard Adam with L2 regularization in modern deep learning?
A
Standard Adam ties L2 regularization to gradient magnitude adaptation, whereas AdamW decouples weight decay, applying it directly to weights and restoring true regularization behavior
✓ Correct
B
AdamW runs twice as fast on TPU hardware
C
AdamW eliminates the need for learning rates
D
AdamW works only on convolutional layers
💡
Step-by-Step Explanation & Concept Rationale
Loshchilov & Hutter showed that in Adam, L2 regularization is distorted by moving average scaling; decoupling weight decay significantly improves generalization.
Q. 9
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Batch Normalization' (BatchNorm) and what are its primary benefits during deep network training?
A
Normalizing layer inputs across the mini-batch to zero mean and unit variance, which stabilizes internal covariate shift, accelerates training, and acts as a mild regularizer
✓ Correct
B
Normalizing the total number of training epochs
C
Rounding all weights to 8-bit integers
D
Scaling the learning rate by the number of GPUs
💡
Step-by-Step Explanation & Concept Rationale
BatchNorm adds learnable scale (gamma) and shift (beta) parameters, enabling higher learning rates and reducing sensitivity to weight initialization.
Q. 10
Artificial Intelligence
Difficulty: Hard
(1 Mark)
Why is 'Layer Normalization' (LayerNorm) universally preferred over Batch Normalization in Transformers and Recurrent Neural Networks (RNNs)?
A
LayerNorm computes mean and variance across the feature/channel dimension for each individual sample independently of batch size, making it ideal for variable-length sequences and batch size = 1
✓ Correct
B
LayerNorm uses GPU shared memory while BatchNorm uses disk storage
C
LayerNorm requires no floating-point math
D
LayerNorm works only on 2D images
💡
Step-by-Step Explanation & Concept Rationale
BatchNorm depends on mini-batch statistics, failing with small batch sizes or dynamic sequence lengths; LayerNorm normalizes across hidden dimensions per token.
Q. 11
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the mechanism and regularization effect of 'Dropout' during neural network training?
A
Randomly setting a fraction p of neuron activations to zero during each forward pass, preventing complex co-adaptations of feature detectors
✓ Correct
B
Permanently deleting 50% of the layers from the model file
C
Dropping slow internet connection packets during distributed training
D
Stopping training early when loss reaches a threshold
💡
Step-by-Step Explanation & Concept Rationale
Dropout forces the network to learn robust, redundant representations by effectively sampling a thinned ensemble of sub-networks.
Q. 12
Artificial Intelligence
Difficulty: Hard
(1 Mark)
During inference/testing, how is Dropout handled?
A
Dropout is disabled (all neurons are active), and activations are scaled by (1 - p) (or inverted dropout scales by 1/(1-p) during training) to preserve expected output magnitude
✓ Correct
B
Dropout rate is increased to 90%
C
All weights are randomly permuted
D
The network computes outputs using only 1 neuron
💡
Step-by-Step Explanation & Concept Rationale
Inverted dropout scales activations during training by 1/(1-p) so that standard unscaled forward passes can execute at test time.
Q. 13
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'He (Kaiming) Initialization' and why is it recommended for neural networks using ReLU activations?
A
Drawing weights from a Gaussian distribution with variance 2 / n_in to keep the variance of activations and backpropagated gradients constant across layers with half-rectified activations
✓ Correct
B
Initializing all weights to exactly 1.0
C
Setting all weights to zero
D
Drawing weights from a uniform distribution between -100 and +100
💡
Step-by-Step Explanation & Concept Rationale
Because ReLU zeroes out negative inputs (halving variance), Kaiming initialization uses a factor of 2 (unlike Xavier/Glorot's 1/n) to prevent signal attenuation.
Q. 14
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What happens if all weights in a multi-layer neural network are initialized to identical constant values (e.g. all zeros or all ones)?
A
All hidden units in a layer compute identical outputs and receive identical gradients, preventing the network from breaking symmetry and learning distinct features
✓ Correct
B
The network trains at maximum theoretical speed
C
The loss function immediately converges to zero
D
The model becomes immune to overfitting
💡
Step-by-Step Explanation & Concept Rationale
Symmetry breaking requires random weight initialization so that different neurons track different input features.
Q. 15
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the core mathematical operation performed by a 2D Convolutional layer in a CNN?
A
Element-wise multiplication of a learnable filter/kernel with localized receptive field patches followed by summation and addition of a bias term
✓ Correct
B
Matrix inversion of the full image matrix
C
Sorting pixels in ascending brightness order
D
Computing the determinant of image RGB channels
💡
Step-by-Step Explanation & Concept Rationale
Convolution exploits spatial locality and translation invariance through shared filter weights sliding across input feature maps.
Q. 16
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Stride' in a convolutional layer?
A
The step size (number of pixels) by which the convolutional filter shifts horizontally and vertically across the input feature map
✓ Correct
B
The thickness of the input image border padding
C
The number of color channels in the image
D
The total number of filters in the layer
💡
Step-by-Step Explanation & Concept Rationale
A stride of 1 shifts the kernel 1 pixel at a time; a stride of 2 skips every other pixel, halving spatial output dimensions (downsampling).
Q. 17
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Padding' (e.g. 'same' padding) in CNNs?
A
Adding artificial pixels (typically zeros) around the spatial borders of an input feature map to preserve spatial dimensions after convolution and retain border information
✓ Correct
B
Compressing image files into JPEG format
C
Normalizing pixel values between 0 and 1
D
Deleting dark pixels from image edges
💡
Step-by-Step Explanation & Concept Rationale
'Same' padding ensures the output spatial dimensions equal the input spatial dimensions (when stride=1); 'valid' padding performs no padding.
Q. 18
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the purpose of a 'Max Pooling' layer in CNN architectures?
A
Downsampling spatial dimensions (height and width) of feature maps, reducing computational parameters while providing spatial translation invariance to minor distortions
✓ Correct
B
Increasing the number of color channels
C
Learning non-linear weights via backpropagation
D
Normalizing pixel brightness across the batch
💡
Step-by-Step Explanation & Concept Rationale
Max pooling extracts the maximum value in local windows (e.g. 2x2 with stride 2), retaining dominant features while halving resolution.
Q. 19
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Global Average Pooling' (GAP) and why is it used to replace dense fully-connected layers at the end of modern CNNs (e.g. ResNet)?
A
Averaging each feature map across its entire spatial dimensions (H x W) into a single scalar, drastically reducing parameter count and preventing overfitting
✓ Correct
B
Averaging pixel values across all images in a dataset
C
Computing the average learning rate across all epochs
D
Normalizing weights to sum to 1
💡
Step-by-Step Explanation & Concept Rationale
GAP converts an (N, C, H, W) tensor directly into (N, C, 1, 1), eliminating millions of parameters associated with flattening into large FC layers.
Q. 20
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What breakthrough architectural mechanism did 'ResNet' (Residual Networks) introduce to successfully train ultra-deep networks (100+ layers)?
A
Skip (Residual / Shortcut) connections that perform identity mapping: H(x) = F(x) + x, allowing gradients to flow directly backward without attenuation
✓ Correct
B
Replacing all convolutions with dense matrix multiplications
C
Using genetic algorithms instead of gradient descent
D
Eliminating activation functions completely
💡
Step-by-Step Explanation & Concept Rationale
Residual connections let layers learn perturbation residuals F(x) around identity mappings, overcoming vanishing gradients in 152-layer networks.
Q. 21
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is a 'Bottleneck Block' in ResNet-50 and deeper ResNet variants?
A
A 3-layer residual block using 1x1 conv (channel reduction), 3x3 conv (spatial processing), and 1x1 conv (channel restoration) to reduce computational FLOPs
✓ Correct
B
A narrow bottleneck in computer RAM data transfer
C
A queue of images waiting for GPU processing
D
A layer that halts gradient backpropagation
💡
Step-by-Step Explanation & Concept Rationale
Bottleneck design cuts computation by compressing channel dimensions before expensive 3x3 convolutions.
Q. 22
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In 'Inception' (GoogLeNet) architectures, what is the role of 1x1 convolutions?
A
Dimensionality reduction (channel pooling) before expensive 3x3 and 5x5 spatial convolutions, drastically reducing computational FLOPs
✓ Correct
B
Upsampling image resolution by 2x
C
Converting RGB images to grayscale
D
Rotating images by 90 degrees
💡
Step-by-Step Explanation & Concept Rationale
1x1 convolutions perform linear combinations across feature channels, shrinking depth while adding non-linear expressiveness.
Q. 23
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What compound scaling method does 'EfficientNet' use to scale network depth, width, and image resolution simultaneously?
A
Compound Coefficient scaling that balances depth (d = alpha^phi), width (w = beta^phi), and resolution (r = gamma^phi) under fixed resource constraints
✓ Correct
B
Arbitrarily doubling network depth until memory runs out
C
Increasing image resolution while keeping depth constant
D
Scaling only the batch size exponentially
💡
Step-by-Step Explanation & Concept Rationale
Tan & Le proved that coordinating depth, width, and resolution scaling yields state-of-the-art accuracy with order-of-magnitude fewer parameters.
Q. 24
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Depthwise Separable Convolution' (used in MobileNet and Xception)?
A
Splitting standard convolution into a Depthwise Convolution (applying 1 spatial filter per input channel) followed by a Pointwise Convolution (1x1 conv combining channels)
✓ Correct
B
Applying convolutions only along the diagonal of an image
C
Convolving across 3D video frames simultaneously
D
Separating red, green, and blue color channels permanently
💡
Step-by-Step Explanation & Concept Rationale
Depthwise separable convolutions reduce computation by ~8x to 9x with minimal loss in accuracy, enabling real-time edge AI on mobile devices.
Q. 25
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the primary limitation of standard Recurrent Neural Networks (vanilla RNNs) on long sequential data?
A
Exploding and vanishing gradients across long unrolled time steps, preventing the network from capturing long-term temporal dependencies
✓ Correct
B
Inability to process text data
C
Requirement for 3D tensor inputs
D
Lack of matrix multiplication operations
💡
Step-by-Step Explanation & Concept Rationale
Repeated multiplication of transition matrix W_hh across hundreds of time steps drives gradients to zero or infinity, creating short-term memory limits.
Study Stream Progress:
Showing 25 of 82 Questions (30%)
Jump to:
🎉
All 82 Solved Questions in Deep Learning Architectures & Optimization Loaded!
You have studied the complete question pool with step-by-step rationales. Ready to test yourself?
Q. 1
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In time-series analysis, what three conditions must be satisfied for a series to be considered 'Weakly Stationary' (Covariance Stationary)?
A
1. Constant mean over time, 2. Constant variance over time, 3. Autocovariance depends only on the time lag between observations and not on actual time t
✓ Correct
B
1. Zero mean, 2. Zero variance, 3. Zero correlation
C
1. Infinite trend, 2. Extreme seasonality, 3. Random noise
D
1. Strictly positive numbers, 2. Strictly integer values, 3. Equal length intervals
💡
Step-by-Step Explanation & Concept Rationale
Stationarity ensures statistical properties do not change over time, a prerequisite for linear forecasting models like ARIMA.
Q. 2
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What statistical hypothesis test is standardly used to test for the presence of a unit root (non-stationarity) in a univariate time series?
A
Augmented Dickey-Fuller (ADF) Test
✓ Correct
B
Student's t-test
C
Chi-Square test of independence
D
ANOVA F-test
💡
Step-by-Step Explanation & Concept Rationale
The ADF test checks the null hypothesis that a unit root is present (non-stationary); p-value < 0.05 rejects the null, indicating stationarity.
Q. 3
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In an ARIMA(p, d, q) model, what do the parameters p, d, and q represent?
A
p = Order of the Autoregressive (AR) terms, d = Degree of differencing (Integration), q = Order of the Moving Average (MA) terms
✓ Correct
B
p = Population, d = Density, q = Quantity
C
p = Precision, d = Dimension, q = Quantile
D
p = Penalty, d = Depth, q = Queue size
💡
Step-by-Step Explanation & Concept Rationale
ARIMA models combine differencing 'd' to achieve stationarity with AR(p) past lags and MA(q) past forecast error terms.
Q. 4
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What visual diagnostic plots are examined to determine the appropriate AR order (p) and MA order (q) in Box-Jenkins ARIMA modeling?
A
Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plots
✓ Correct
B
ROC curve and Precision-Recall curve
C
Confusion matrix and Silhouette plot
D
Scatter plot and Box plot
💡
Step-by-Step Explanation & Concept Rationale
PACF cuts off after lag p for AR(p) processes; ACF cuts off after lag q for MA(q) processes.
Q. 5
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'SARIMA' (Seasonal ARIMA: ARIMA(p,d,q)(P,D,Q)_s) and what does parameter 's' signify?
A
Seasonal ARIMA incorporating seasonal AR, differencing, and MA terms; 's' is the seasonal period length (e.g. s=12 for monthly data, s=24 for hourly)
✓ Correct
B
A model for predicting summer temperatures only
C
A software tool for running ARIMA on GPUs
D
A method that ignores seasonal cycles completely
💡
Step-by-Step Explanation & Concept Rationale
SARIMA captures both non-seasonal short-term dynamics and recurring seasonal cycle patterns (e.g. annual monsoon rainfall peaks).
Q. 6
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Prophet' (developed by Meta / Facebook) additive time-series forecasting model?
A
A decomposable time-series model: y(t) = g(t) + s(t) + h(t) + epsilon_t, combining non-linear trends with changepoint detection, seasonality (Fourier series), and holiday effects
✓ Correct
B
A religious astrology software
C
A deep learning vision model for satellite photos
D
A recurrent neural network for natural language text
💡
Step-by-Step Explanation & Concept Rationale
Taylor & Letham designed Prophet to handle real-world business and environmental time series with missing data, trend shifts, and robust outlier handling.
Q. 7
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'DeepAR' (Salinas et al. / Amazon Research) for probabilistic time-series forecasting?
A
An autoregressive recurrent neural network (LSTM/GRU) that learns global non-linear representations across thousands of related time series, outputting parametric probability distributions
✓ Correct
B
A model that only forecasts deep ocean water currents
C
A classical linear regression equation
D
A decision tree model for tabular classification
💡
Step-by-Step Explanation & Concept Rationale
DeepAR trains across large item catalogs or multi-sensor networks (e.g. 500 river gauge stations), predicting probabilistic quantile forecasts.
Q. 8
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Temporal Fusion Transformer' (TFT, Lim et al. / Google Cloud AI)?
A
An attention-based architecture for multi-horizon time-series forecasting that combines Variable Selection Networks, Gated Residual Networks, and Multi-Head Self-Attention with temporal interpretability
✓ Correct
B
A transformer that alters time in video games
C
A high-voltage electrical power transformer
D
A software tool for scheduling calendar meetings
💡
Step-by-Step Explanation & Concept Rationale
TFT handles heterogeneous inputs (static metadata, known future inputs like rain forecasts, and observed past histories) with native prediction intervals.
Q. 9
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'N-BEATS' (Neural Basis Expansion Analysis for Interpretable Time Series, Oreshkin et al.)?
A
A deep neural architecture composed of backward and forward residual link blocks using basis expansions to perform pure deep learning time-series forecasting without recurrent or attention layers
✓ Correct
B
A music rhythm generation software
C
A heartbeat monitoring device
D
A video streaming compression tool
💡
Step-by-Step Explanation & Concept Rationale
N-BEATS surpassed classical statistical ensembles in the M4 competition by decomposing forecasts into interpretable trend and seasonality basis functions.
Q. 10
Artificial Intelligence
Difficulty: Hard
(1 Mark)
When using LSTM networks for multi-step river flood discharge forecasting, what is 'Autoregressive Multi-Step' versus 'Direct Multi-Step' forecasting?
A
Autoregressive iteratively feeds single-step predictions back as inputs for subsequent steps; Direct trains separate independent models for each future forecast horizon (t+1, t+2, ..., t+H)
✓ Correct
B
Autoregressive runs on CPU; Direct runs on GPU
C
Autoregressive uses images; Direct uses audio
D
There is no difference in their forecasting architectures
💡
Step-by-Step Explanation & Concept Rationale
Direct forecasting avoids cascading error accumulation inherent in autoregressive rollouts at the cost of training H independent models.
Q. 11
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'MLflow' (open-source MLOps platform) and what are its 4 core components?
A
1. MLflow Tracking (logging experiments, metrics, parameters), 2. MLflow Projects (packaging code), 3. MLflow Models (deploying models), 4. MLflow Model Registry (centralized model lifecycle management)
✓ Correct
B
1. Monitor, 2. Keyboard, 3. Mouse, 4. Processor
C
1. CPU, 2. GPU, 3. RAM, 4. Hard Drive
D
1. Login, 2. Password, 3. Email, 4. Phone
💡
Step-by-Step Explanation & Concept Rationale
MLflow manages the complete machine learning lifecycle from experimentation to model registry and deployment.
Q. 12
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'DVC' (Data Version Control) in modern MLOps pipelines?
A
An open-source Git-extension tool that versions massive datasets, model weights, and intermediate pipeline artifacts using content-addressable storage (S3, GCS, local storage) without bloating Git repositories
✓ Correct
B
A digital video camera recording software
C
A driver verification card for trucks
D
A database management system for SQL tables
💡
Step-by-Step Explanation & Concept Rationale
DVC stores lightweight .dvc pointer files in Git while syncing large multi-gigabyte data files and model checkpoints to remote cloud buckets.
Q. 13
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Docker' containerization and why is it essential for machine learning reproducibility and deployment?
A
Packaging an ML model along with its code, runtime dependencies, system libraries, CUDA drivers, and Python environment into an isolated, portable container image that runs identically on any infrastructure
✓ Correct
B
A physical shipping container used on cargo vessels
C
A computer monitor calibration tool
D
A file compression utility like WinRAR
💡
Step-by-Step Explanation & Concept Rationale
Containers eliminate 'it works on my machine' dependency conflicts between local development laptops and production cloud clusters.
Q. 14
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Kubernetes' (K8s) in production machine learning infrastructure?
A
An open-source container orchestration platform that automates the deployment, horizontal scaling, load balancing, health monitoring, and management of containerized ML services across server clusters
✓ Correct
B
A programming language developed by Google
C
A database management tool for NoSQL data
D
A graphical design software for websites
💡
Step-by-Step Explanation & Concept Rationale
K8s manages auto-scaling of inference pods (e.g. via HPA based on GPU utilization or request queues) and zero-downtime rolling updates.
Q. 15
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'KServe' (formerly KFServing) on Kubernetes?
A
A cloud-native model serving framework on Kubernetes providing serverless auto-scaling (including scale-to-zero), canary rollouts, and standardized inference protocol for PyTorch, TensorFlow, ONNX, and XGBoost
✓ Correct
B
A customer service ticketing software
C
A restaurant food ordering application
D
A file transfer protocol for servers
💡
Step-by-Step Explanation & Concept Rationale
KServe provides production ML inference capabilities like payload logging, drift detection integration, and multi-model serving out-of-the-box.
Q. 16
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'ONNX' (Open Neural Network Exchange) runtime format?
A
An open, cross-platform standard representation format for machine learning models that allows models trained in PyTorch/TensorFlow to be executed on optimized C++ inference runtimes across diverse hardware
✓ Correct
B
An online video streaming codec
C
An audio format like MP3
D
A database query language
💡
Step-by-Step Explanation & Concept Rationale
ONNX decouples training frameworks from production serving environments, enabling hardware acceleration via ONNX Runtime on Intel, AMD, ARM, and NVIDIA chips.
Q. 17
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Triton Inference Server' (developed by NVIDIA)?
A
An open-source, enterprise-grade multi-framework inference serving software that optimizes GPU/CPU utilization through dynamic batching, concurrent model execution, and shared memory pipelines
✓ Correct
B
A gaming console manufactured by NVIDIA
C
A submarine exploration vehicle
D
A graphics design software suite
💡
Step-by-Step Explanation & Concept Rationale
Triton serves PyTorch, TensorRT, ONNX, and Python models simultaneously with microsecond latency and hardware saturation.
Q. 18
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Dynamic Batching' in Triton Inference Server / vLLM?
A
Automatically grouping independent client inference requests arriving within a small time window (e.g. 2 milliseconds) into a single batch on the GPU to maximize hardware throughput without violating latency SLAs
✓ Correct
B
Batching requests once every 24 hours
C
Combining all training datasets into one file
D
Shuffling data randomly during training
💡
Step-by-Step Explanation & Concept Rationale
Dynamic batching dramatically increases server throughput (QPS) on concurrent traffic while maintaining low response latency for individual users.
Q. 19
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'FastAPI' and why is it preferred over Flask for building production Python ML APIs?
A
FastAPI is built on Starlette and Pydantic, supporting native asynchronous execution (async/await), automatic request data validation, high throughput, and automatic OpenAPI (Swagger) documentation generation
✓ Correct
B
FastAPI only works with fast computers
C
FastAPI requires no programming knowledge
D
FastAPI executes on GPU hardware only
💡
Step-by-Step Explanation & Concept Rationale
FastAPI delivers NodeJS/Go-level performance in Python and simplifies input/output schema validation for complex ML payloads.
Q. 20
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'gRPC' compared to REST (HTTP/JSON) for high-performance ML inference microservices?
A
gRPC uses HTTP/2 with binary Protocol Buffers (Protobuf) serialization, delivering substantially lower serialization overhead, lower network latency, and bidirectional streaming compared to text-based JSON over HTTP/1.1
✓ Correct
B
gRPC is for web browsers only; REST is for servers
C
gRPC requires no internet connection
D
There is no performance difference between them
💡
Step-by-Step Explanation & Concept Rationale
In high-throughput computer vision and audio pipelines, Protobuf serialization in gRPC avoids JSON text parsing bottlenecks.
Q. 21
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Model Drift' (Model Decay) in production ML systems?
A
The gradual degradation of a model's predictive performance over time caused by evolving real-world conditions, changing user behaviors, or unaligned data distributions
✓ Correct
B
The physical movement of a computer server on a table
C
The loss of electricity to a server room
D
A bug in Python code that deletes files
💡
Step-by-Step Explanation & Concept Rationale
Model drift requires continuous telemetry monitoring and automated retraining triggers to maintain production accuracy.
Q. 22
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Evidently AI' / 'Whylogs' library used for in MLOps?
A
Continuous monitoring and visualization of data drift, target drift, model quality metrics, and feature distribution shifts in production ML pipelines
✓ Correct
B
Building 3D video games
C
Editing audio podcasts
D
Formatting source code indentation
💡
Step-by-Step Explanation & Concept Rationale
Evidently computes statistical distance tests (KS test, Wasserstein distance, PSI) to generate production drift dashboards.
Q. 23
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Feature Store Online vs Offline Store' architecture?
A
Offline store (Parquet/Snowflake/BigQuery) is optimized for high-throughput batch historical feature retrieval for training; Online store (Redis/DynamoDB) is optimized for ultra-low latency (<10ms) single-record feature lookups for inference
✓ Correct
B
Offline is disconnected from power; Online is connected
C
Offline is for images; Online is for text
D
There is no architectural difference between them
💡
Step-by-Step Explanation & Concept Rationale
Dual-store feature architecture guarantees feature consistency while satisfying opposing throughput and latency requirements.
Q. 24
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'CI/CD for Machine Learning' (CD4ML / MLOps Pipeline)?
A
Continuous Integration, Continuous Delivery, and Continuous Training practices that automate testing of code, data validation, automated model retraining, and zero-downtime deployment
✓ Correct
B
Writing code without testing
C
Deploying models manually by copying files via USB drive
D
Shutting down servers during software updates
💡
Step-by-Step Explanation & Concept Rationale
CD4ML treats data and model pipelines with the same automated rigor as traditional software engineering.
Q. 25
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Great Expectations' tool in data quality engineering?
A
A Python data validation framework that asserts, documents, and monitors data pipeline expectations (e.g. column ranges, nullness, types) before data enters model training
✓ Correct
B
A novel by Charles Dickens
C
A philosophical theory on AI ethics
D
A performance rating given to employees
💡
Step-by-Step Explanation & Concept Rationale
Great Expectations catches corrupt, out-of-bounds, or incomplete data at ingestion checkpoints, preventing silent model training failures.
Study Stream Progress:
Showing 25 of 82 Questions (30%)
Jump to:
🎉
All 82 Solved Questions in Time-Series Forecasting & MLOps Deployment Loaded!
You have studied the complete question pool with step-by-step rationales. Ready to test yourself?
Q. 1
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is the core mathematical formula for Scaled Dot-Product Attention in the original Transformer ('Attention Is All You Need', Vaswani et al. 2017)?
A
Attention(Q, K, V) = softmax((Q * K^T) / sqrt(d_k)) * V
✓ Correct
B
Attention(Q, K, V) = sigmoid(Q * K + V)
C
Attention(Q, K, V) = tanh((Q + K) * V)
D
Attention(Q, K, V) = ReLU(Q * K^T) * V
💡
Step-by-Step Explanation & Concept Rationale
Dividing by sqrt(d_k) prevents dot products from growing excessively large for large dimensions, which would otherwise push the softmax into regions with vanishing gradients.
Q. 2
Artificial Intelligence
Difficulty: Hard
(1 Mark)
Why is the scaling factor 1 / sqrt(d_k) crucial in Scaled Dot-Product Attention?
A
For large key dimensions d_k, the dot products grow large in magnitude, pushing the softmax function into regions with extremely small gradients
✓ Correct
B
It reduces matrix memory size on GPU
C
It converts complex numbers to real floats
D
It bounds output values strictly between -1 and 1
💡
Step-by-Step Explanation & Concept Rationale
Under independent zero-mean unit-variance components, the dot product has mean 0 and variance d_k; dividing by sqrt(d_k) restores unit variance.
Q. 3
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In Multi-Head Attention, why are multiple projection heads used rather than a single large attention head?
A
Multiple heads allow the model to jointly attend to information from different representation subspaces at different positions simultaneously
✓ Correct
B
Multiple heads allow the model to train on multiple GPUs without code changes
C
Multiple heads reduce the total number of matrix multiplications
D
Multiple heads eliminate the need for feed-forward layers
💡
Step-by-Step Explanation & Concept Rationale
Multi-Head Attention: Concat(head_1, ..., head_h) * W_O, where each head can capture distinct syntactic and semantic relationships.
Q. 4
Artificial Intelligence
Difficulty: Medium
(1 Mark)
Why are 'Positional Encodings' strictly necessary in Transformer architectures?
A
Because the self-attention mechanism is inherently permutation-equivariant (it has no built-in notion of token order or sequence position)
✓ Correct
B
To compress long sentences into short embeddings
C
To translate text between different natural languages
D
To prevent integer overflow during softmax calculation
💡
Step-by-Step Explanation & Concept Rationale
Unlike RNNs which process tokens step-by-step, Transformers process all tokens in parallel; without positional encodings, word order is lost.
Q. 5
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Byte-Pair Encoding' (BPE) tokenization widely used in GPT models?
A
A subword tokenization algorithm that iteratively merges the most frequent pair of adjacent bytes or characters in a text corpus into a new vocabulary token
✓ Correct
B
An encryption algorithm that secures text against cyberattacks
C
A method that converts text into raw ASCII binary bits
D
A tokenization method that splits text strictly on whitespaces
💡
Step-by-Step Explanation & Concept Rationale
BPE balances word-level semantics with character-level coverage, eliminating Out-Of-Vocabulary (OOV) tokens by decomposing rare words into subword pieces.
Q. 6
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'WordPiece' tokenization (used in BERT)?
A
A subword tokenization algorithm that selects symbol pairs to merge based on maximizing the likelihood of the training language model rather than pure frequency count
✓ Correct
B
A tool for checking spelling mistakes in Microsoft Word
C
A method that limits vocabularies to 1,000 words only
D
A syntax parser that extracts noun phrases
💡
Step-by-Step Explanation & Concept Rationale
WordPiece uses the '##' prefix for subword continuations (e.g. 'playing' -> 'play', '##ing') and optimizes corpus likelihood.
Q. 7
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'SentencePiece' tokenization (used in T5, LLaMA)?
A
A language-independent subword tokenizer that treats raw input text as a stream of Unicode characters including whitespace (represented as '_'), without requiring language-specific pre-tokenizers
✓ Correct
B
A model that generates full sentences from single words
C
A grammar checking rule engine
D
A tool that counts the number of sentences in a document
💡
Step-by-Step Explanation & Concept Rationale
SentencePiece enables seamless multi-lingual tokenization without language-specific regex or whitespace assumptions.
Q. 8
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the fundamental architectural difference between BERT and GPT?
A
BERT is an Encoder-only model trained with bidirectional masked language modeling; GPT is a Decoder-only autoregressive model trained with causal left-to-right next-token prediction
✓ Correct
B
BERT is for images; GPT is for audio
C
BERT has no attention layers; GPT uses only attention layers
D
BERT runs only on CPUs; GPT runs on TPUs
💡
Step-by-Step Explanation & Concept Rationale
BERT excels at representation and comprehension (classification, NER, QA); GPT excels at generative text synthesis.
Q. 9
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Masked Language Modeling' (MLM) pre-training objective introduced by BERT?
A
Randomly masking 15% of input tokens with a [MASK] token and training the bidirectional encoder to predict the original identity of the masked words from context
✓ Correct
B
Hiding user personal data for privacy
C
Masking convolutional feature maps with zero padding
D
Filtering out offensive words from internet text
💡
Step-by-Step Explanation & Concept Rationale
MLM enables deep bidirectional representation learning by forcing the model to fuse left and right context representations simultaneously.
Q. 10
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'Causal Masking' (Autoregressive Masking / Look-Ahead Mask) in Transformer Decoders (e.g. GPT)?
A
An upper-triangular attention mask that sets future token attention scores to -infinity before softmax, ensuring position i can only attend to positions <= i
✓ Correct
B
A mask that prevents models from generating offensive text
C
A mask that hides images from text decoders
D
A technique that deletes punctuation tokens
💡
Step-by-Step Explanation & Concept Rationale
Causal masking enforces the autoregressive property during parallel training: token t cannot cheat by looking at ground-truth future token t+1.
Q. 11
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'RoBERTa' (Robustly Optimized BERT Approach) and what key modifications did it make to BERT?
A
Trained longer on larger batches with more data, removed the Next Sentence Prediction (NSP) task, used dynamic masking, and adopted a larger byte-level BPE vocabulary
✓ Correct
B
Replaced all self-attention with convolutional filters
C
Reduced the number of layers to 2
D
Added a robotic text-to-speech module
💡
Step-by-Step Explanation & Concept Rationale
Liu et al. proved BERT was significantly undertrained; removing NSP and scaling data/batch size dramatically boosted GLUE benchmark scores.
Q. 12
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is the 'T5' (Text-to-Text Transfer Transformer) framework introduced by Google?
A
A unified Encoder-Decoder model that casts every NLP problem (classification, summarization, translation, QA) into a text-to-text format
✓ Correct
B
A 5-layer convolutional vision model
C
A programming language for training neural nets
D
A tool for converting text into SQL databases
💡
Step-by-Step Explanation & Concept Rationale
T5 uses prefix prompts (e.g. 'translate English to German: ...') to handle diverse NLP tasks within a single unified architecture.
Q. 13
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In Retrieval-Augmented Generation (RAG), what is the primary workflow?
A
1. Chunk and embed private documents into a Vector Database, 2. Retrieve top-k relevant chunks based on user query similarity, 3. Augment the LLM prompt with retrieved context for grounded generation
✓ Correct
B
1. Retrain the entire LLM from scratch on private data every day
C
1. Randomly guess relevant facts from Wikipedia
D
1. Convert text documents into audio files
💡
Step-by-Step Explanation & Concept Rationale
RAG overcomes LLM hallucinations and knowledge cutoff dates by dynamically retrieving factual domain context at inference time.
Q. 14
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is a 'Vector Database' (e.g. Pinecone, Milvus, Qdrant, Chroma, pgvector) optimized for?
A
High-throughput, low-latency Approximate Nearest Neighbor (ANN) search over high-dimensional dense embedding vectors (e.g. HNSW, IVF-PQ)
✓ Correct
B
Storing relational SQL tables with primary keys only
C
Storing raw video files for streaming
D
Compiling C++ programs into machine code
💡
Step-by-Step Explanation & Concept Rationale
Vector databases index dense vector embeddings to perform sub-millisecond similarity queries across millions of multi-dimensional vectors.
Q. 15
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'HNSW' (Hierarchical Navigable Small World) in vector search indexing?
A
A multi-layer graph-based index for fast Approximate Nearest Neighbor (ANN) search with logarithmic O(log N) search complexity
✓ Correct
B
A hardware acceleration chip for GPUs
C
A network security firewall protocol
D
A loss function for training BERT models
💡
Step-by-Step Explanation & Concept Rationale
HNSW builds a hierarchy of proximity graphs where upper layers enable fast coarse routing and bottom layers execute precise local neighbor search.
Q. 16
Artificial Intelligence
Difficulty: Hard
(1 Mark)
In dense text retrieval, what is 'Bi-Encoder' versus 'Cross-Encoder' architecture?
A
Bi-Encoders encode query and document independently into vectors for fast vector DB search; Cross-Encoders process (query, doc) pairs jointly through full self-attention for higher accuracy reranking
✓ Correct
B
Bi-Encoders use two GPUs; Cross-Encoders use one GPU
C
Bi-Encoders are for images; Cross-Encoders are for audio
D
There is no difference between them
💡
Step-by-Step Explanation & Concept Rationale
Bi-encoders (e.g. sentence-transformers) enable sub-millisecond vector retrieval; cross-encoders are used as second-stage rerankers on top-100 candidates.
Q. 17
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'QLoRA' (Quantized Low-Rank Adaptation) and how does it enable fine-tuning a 65B LLM on a single 48GB GPU?
A
Quantizing the base model weights to 4-bit NormalFloat (NF4), using Double Quantization to compress quantization constants, and applying Paged Optimizers to manage memory spikes during LoRA fine-tuning
✓ Correct
B
Deleting 75% of the base model weights permanently
C
Training only the first 2 layers of the network
D
Replacing all attention layers with linear regression
💡
Step-by-Step Explanation & Concept Rationale
Dettmers et al. proved QLoRA matches full 16-bit fine-tuning performance while reducing required GPU VRAM from >780GB to <48GB.
Q. 18
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'RLHF' (Reinforcement Learning from Human Feedback) in LLM alignment?
A
1. Train a Reward Model on human preference pairwise comparisons, 2. Fine-tune the base policy LLM using PPO (Proximal Policy Optimization) against the reward model with a KL penalty
✓ Correct
B
Asking humans to write all LLM responses manually in real-time
C
Training models by punishing developers when errors occur
D
A game where humans play chess against AI models
💡
Step-by-Step Explanation & Concept Rationale
RLHF aligns raw pre-trained LLMs to be helpful, honest, and harmless (HHH) based on human preference signals.
Q. 19
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'DPO' (Direct Preference Optimization) and how does it simplify LLM alignment compared to RLHF with PPO?
A
It mathematically reparameterizes the reward model directly in terms of the language model policy, optimizing preference data with a simple closed-form cross-entropy loss without training a separate reward model or using unstable RL (PPO)
✓ Correct
B
It replaces text tokens with numerical integers
C
It trains models with direct supervised classification labels only
D
It uses genetic evolution instead of neural nets
💡
Step-by-Step Explanation & Concept Rationale
Rafailov et al. showed DPO eliminates the complexity and instability of PPO while achieving equal or superior alignment performance.
Q. 20
Artificial Intelligence
Difficulty: Medium
(1 Mark)
In prompt engineering, what is 'Chain-of-Thought' (CoT) prompting (Wei et al. 2022)?
A
Prompting the LLM to generate explicit intermediate reasoning steps before arriving at the final answer ('Let's think step by step'), dramatically improving multi-step mathematical and logical reasoning
✓ Correct
B
Connecting multiple LLMs with physical cables
C
Sending prompts in a continuous loop to overload the server
D
Translating prompts through 5 different languages
💡
Step-by-Step Explanation & Concept Rationale
CoT allocates more test-time computation to intermediate token generation, allowing models to decompose complex reasoning problems.
Q. 21
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Self-Consistency' in Chain-of-Thought prompting?
A
Sampling multiple diverse reasoning paths from the LLM with temperature > 0 and selecting the final answer via majority vote across all sampled paths
✓ Correct
B
Ensuring the LLM never changes its answer
C
Validating Python code syntax
D
Checking that prompt words are spelled correctly
💡
Step-by-Step Explanation & Concept Rationale
Wang et al. showed self-consistency over multiple reasoning trajectories delivers significant reasoning accuracy gains.
Q. 22
Artificial Intelligence
Difficulty: Hard
(1 Mark)
What is 'ReAct' (Reasoning + Acting) framework for LLM autonomous agents?
A
An agentic paradigm that interleaves reasoning traces ('Thought'), action execution ('Action' via external tools/APIs), and environmental feedback ('Observation')
✓ Correct
B
A JavaScript library for building web UIs
C
A chemical reaction simulation engine
D
A reactive programming library for mobile apps
💡
Step-by-Step Explanation & Concept Rationale
Yao et al. demonstrated that combining reasoning with external tool usage (search, Python execution, SQL) enables robust problem solving.
Q. 23
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Tool Calling' (Function Calling) in modern LLM APIs?
A
The capability of an LLM to detect when an external function should be called, outputting structured JSON arguments matching a predefined schema rather than plain text
✓ Correct
B
Making telephone calls using voice AI
C
Calling Python functions manually from terminal
D
Installing development tools on Linux
💡
Step-by-Step Explanation & Concept Rationale
Function calling enables deterministic integration between LLMs and databases, APIs, calculation engines, and ERP systems.
Q. 24
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'Hallucination' in Large Language Models and what are its primary causes?
A
The generation of factually incorrect, fabricated, or nonsensical assertions presented with high confidence, caused by statistical pattern completion over noisy training data and lack of grounded world state
✓ Correct
B
A visual display glitch on GPU monitors
C
A virus infecting the LLM server
D
A power outage during inference
💡
Step-by-Step Explanation & Concept Rationale
LLMs optimize for plausible sequence continuation rather than absolute factual truth; mitigated via RAG, citation grounding, and guardrails.
Q. 25
Artificial Intelligence
Difficulty: Medium
(1 Mark)
What is 'BLEU' (Bilingual Evaluation Understudy) score and what is its primary use?
A
A metric measuring the n-gram overlap precision between machine-generated translations and human reference translations (with a brevity penalty for short outputs)
✓ Correct
B
A metric measuring image brightness
C
A test for measuring internet bandwidth
D
A tool for grading spoken pronunciation
💡
Step-by-Step Explanation & Concept Rationale
BLEU computes geometric average of n-gram precisions (n=1 to 4) multiplied by brevity penalty BP.
Study Stream Progress:
Showing 25 of 82 Questions (30%)
Jump to:
🎉
All 82 Solved Questions in Transformers, LLMs & Natural Language Processing Loaded!
You have studied the complete question pool with step-by-step rationales. Ready to test yourself?
Ready to Test Your Retention & Speed?
Now that you have reviewed the study questions and rationales, test yourself in our interactive 1-by-1 practice engine or take the full official timed mock exam.