1
© 2025 AIML MCQ
AI & Machine Learning MCQ
63+ Multiple Choice Questions with Detailed Explanations
Practice questions on Machine Learning, Deep Learning, RNN, and Neural Networks
All Questions (Mixed)
63 Questions2
What is a convolutional neural network (CNN) primarily used for?
3
What is a Recurrent Neural Network (RNN) best suited for?
4
What is a perceptron?
5
Which algorithm is used for classification problems?
6
What is the purpose of the activation function in neural networks?
7
What problem do LSTM networks solve?
8
What is the ReLU activation function?
9
What is overfitting in machine learning?
10
What is backpropagation?
11
What are the three gates in an LSTM cell?
12
What is the purpose of the softmax function?
13
Which technique is used to prevent overfitting?
14
What is dropout in deep learning?
15
What is a GRU (Gated Recurrent Unit)?
16
What is weight initialization?
17
What does the bias-variance tradeoff represent?
18
What is a pooling layer in CNN?
19
What is bidirectional RNN?
20
What is a hidden layer in a neural network?
21
What is cross-validation used for?
22
What is transfer learning?
23
What is teacher forcing in RNN training?
24
What is the loss function used for?
25
Which is an example of unsupervised learning?
26
What is batch normalization?
27
What is sequence-to-sequence (Seq2Seq) modeling?
28
What is gradient descent?
29
What is the purpose of feature scaling?
30
Which optimizer is commonly used in deep learning?
31
What is the attention mechanism in RNNs?
32
What is the learning rate in neural networks?
33
What is a decision tree?
34
What is the vanishing gradient problem?
35
What is a common application of RNNs?
36
What is an epoch in neural network training?
37
What is ensemble learning?
38
What is a GAN (Generative Adversarial Network)?
39
What is the exploding gradient problem in RNNs?
40
What is a fully connected layer?
41
In the context of Support Vector Machines (SVM), what is the purpose of the kernel trick?
42
In modern CNN architectures like Inception and ResNet, what is the primary purpose of a 1x1 Convolution?
43
What is the purpose of Truncated Backpropagation Through Time (TBPTT) in training Recurrent Neural Networks?
44
What is the primary cause of the "Dying ReLU" problem in neural networks?
45
Consider a dataset with highly imbalanced classes (e.g., 99% Class A, 1% Class B). Which evaluation metric is the LEAST informative for assessing model performance?
46
How does a Batch Normalization layer behave differently during inference (testing) compared to training?
47
Which of the following describes the "Exposure Bias" problem in standard sequence-to-sequence RNN models trained with Teacher Forcing?
48
Why is He Initialization (Kaiming Initialization) preferred over Xavier (Glorot) Initialization for deep networks using ReLU activations?
49
In Principal Component Analysis (PCA), what do the eigenvectors of the covariance matrix represent?
50
What is the critical mathematical mechanism that allows ResNets (Residual Networks) to train networks with over 100 layers without suffering from the vanishing gradient problem?
51
In a Gated Recurrent Unit (GRU), what is the specific role of the Reset Gate?
52
In the context of the Softmax function often used in Knowledge Distillation, what effect does increasing the "Temperature" (T > 1) have on the output probability distribution?
53
Why does L1 regularization (Lasso) tend to produce sparse models compared to L2 regularization (Ridge)?
54
What is the core difference between Semantic Segmentation and Instance Segmentation in computer vision?
55
How does the Attention Mechanism in an RNN-based Seq2Seq model solve the "bottleneck" problem of standard Encoder-Decoder architectures?
56
When analyzing the high-dimensional loss landscapes of deep neural networks, what type of critical point is far more common and problematic than a poor local minimum?
57
In the context of ensemble methods, how does XGBoost fundamentally differ from Random Forest?
58
If a Dropout layer with a dropout rate of 0.5 is applied during training, what mathematical adjustment must happen during inference?
59
While Self-Attention (as seen in Transformers) has largely replaced RNNs in NLP, what fundamental capability does an RNN possess that standard Self-Attention lacks without modification?
60
What is the primary benefit of using a Learning Rate Schedule with Warmup (e.g., Cosine Annealing with Warmup)?
61
When diagnosing a severe disease where early detection is critical, which evaluation threshold tuning strategy is most appropriate?
62
What is the primary advantage of the K-Means++ initialization algorithm over standard K-Means random initialization?
63