GATE DA Machine Learning – Full Mock Test-1 ⋆ CSE Study247

GATE DA Machine Learning – Full Mock Test-1

Questions 91-95 of 100

0% completed (0/100 answered)

Question 91

MSQ · Multiple Correct 2 mark

Which of the following activation functions suffer from the VANISHING GRADIENT problem for large magnitude inputs?

A Sigmoid B Hyperbolic Tangent (Tanh) C Standard ReLU (for z > 0) D Leaky ReLU (for z > 0)

Question 92

MSQ · Multiple Correct 2 mark

Which of the following statements regarding BATCH NORMALIZATION are CORRECT?

A Normalizes activations across the mini-batch dimension for each feature. B Reduces internal covariate shift and stabilizes deep network training. C Introduces learnable scale \gamma and shift $\beta parameters$. D During test-time inference, uses running population mean and variance estimated during training.

Question 93

MSQ · Multiple Correct 2 mark

Which of the following statements regarding ADAM OPTIMIZER are CORRECT?

A Combines concepts from Momentum and RMSprop. B Maintains exponential moving averages of past gradients (1st moment) and past squared gradients (2nd moment). C Includes bias correction steps to compensate for zero initialization of moments. D Requires no learning rate parameter.

Question 94

MSQ · Multiple Correct 2 mark

Which of the following are REGULARIZATION techniques used to prevent overfitting in neural networks?

A Dropout B L2 Weight Decay C Early Stopping D Data Augmentation

Question 95

MSQ · Multiple Correct 2 mark

Which of the following properties hold for the RECTIFIED LINEAR UNIT (ReLU) activation function f(z) = max(0, z)?

A It is computationally efficient (simple thresholding at 0). B Its derivative is 1 for z > 0 and 0 for z < 0. C It is non-differentiable at z = 0. D It is bounded between -1 and +1.