DL_unit-2

Complete Roll Number

What is the primary purpose of the Sigmoid activation function in artificial neural networks?

• Introduce non-linearity

• Enhance computational efficiency

• Guarantee vanishing gradient

• • Eliminate overfitting

Which hyperbolic function is commonly used as an activation function in neural networks?

• Sine

• Cosine

• Hyperbolic Tangent (tanh)

• Hyperbolic Secant (sech)

What is the primary purpose of the Perceptron Training Rule in neural networks?

• Adjust weights to minimize errors

• Control learning rate

• Prevent overfitting

• Guarantee non-linearity

Which activation function is more resilient to the vanishing gradient problem and is commonly used in the hidden layers of deep neural networks?

• Sigmoid

• Hyperbolic Tangent (tanh)

• Rectified Linear Unit

• Softmax

What optimization algorithm is commonly used to train artificial neural networks, and how does it address the task of adjusting weights to minimize errors during training?

• Adversarial training, by adjusting weights based on the gradient of the loss function.

• Genetic algorithms, by evolving a population of networks over generations.

• Gradient Descent, by iteratively updating weights in the direction of steepest descent of the loss function.

• Reinforcement learning, by using reward signals to adjust weights.

Explain the concept of the "Perceptron Training Rule" and its significance in adjusting weights during the training of a single-layer perceptron.

• The rule for adjusting weights based on the magnitude of the input features.

• A strategy for handling imbalanced datasets during training.

• The iterative process of updating weights to minimize errors and improve model accuracy.

• A method for introducing randomness in weight adjustments to prevent overfitting.

The impact of the learning rate in the Gradient Descent Rule for training artificial neural networks. How does an inappropriate learning rate affect the training process, and what strategies can be employed to address this issue?

• A high learning rate may cause oscillations or divergence, while a low learning rate may result in slow convergence. Techniques like learning rate schedules or adaptive learning rates can help address these challenges.

• The learning rate has minimal impact on the training process, and a fixed learning rate is universally suitable for all networks.

• A high learning rate guarantees rapid convergence, while a low learning rate is necessary for stable training. Adjusting the learning rate dynamically is discouraged.

• The learning rate influences only the convergence speed but has no impact on the stability of the training process.

What issue does the vanishing gradient problem pose in deep neural networks, particularly during training?

• Rapid convergence

• Slow convergence

• Slow convergence

• Exploding gradient

What is the primary function of the Softmax activation in neural networks?

• Introduce non-linearity

• Normalize output into probability distribution

• Prevent vanishing gradient

• Eliminate underfitting