Intro to AI Final Exam Review

What is the role of the discriminator_loss and generator_loss variables in a typical GAN training loop?

• They are hyperparameters that control the learning rate.

• They are used to track the accuracy of the discriminator and generator.

• They store the model weights of the discriminator and generator respectively.

• They represent the loss values to be optimized during training.

Which optimizer is commonly used for training GANs in TensorFlow?

• Adagrad

• RMSprop

• SGD (Stochastic Gradient Descent)

• Adam

What is the purpose of the tf.GradientTape context in TensorFlow when training a GAN?

• It visualizes the training progress.

• It automatically updates the model weights.

• It saves the model checkpoints during training.

• It records operations for automatic diff erentiation

Which loss function is commonly used to train the discriminator in a GAN?

• Kullback-Leibler Divergence (KLD)

• Categorical Cross-Entropy (CCE)

• Binary Cross-Entropy (BCE)

• Mean Squared Error (MSE)

Which of the following best describes the relationship between the generator and discriminator in a GAN?

• The generator and discriminator cooperate to produce high-quality samples.

• The generator and discriminator are independent networks with separate objectives.

• The generator tries to maximize the discriminator's loss, while the discriminator tries to minimize its loss.

• The generator tries to minimize the discriminator's loss, while the discriminator tries to maximize its loss.

If the deep learning neural network has the problem of gradient disappearance or gradient explosion, our common solutions are: (Multiple answers)

• Regularization

• Gradient shear

• Random under-sampling

• Use Relu Activation function

What are the regularizations in deep learning? (Multiple answers)

• Dropout

• Data set enhancement

• Integration method

• L1 norm and L2 norm

Which of the following is not part of the deep learning development framework?

• MXNet

• CNTK

• Keras

• SAFA

Which is not a deep learning algorithm?

• Support vector machine

• Self-encoder

• Convolutional neural networks

• Recurrent neural networks

Which of the following is not an optimization method in deep learning?

• Momentum

• Random gradient descent

• Back propagation algorithm

• Principal component analysis