AI QUIZ

AI algorithms perform better when trained with more data

No single algorithm is best for all problems

AI algorithms cannot surpass human intelligence

Training AI models requires computational optimization

Overfitting to the training data

Deletion of weights during training

Loss of previously learned information when learning new tasks

Memory overflow during backpropagation

To reduce training time

To explore the environment

To stabilize learning by fixing the Q-value targets

To store policy gradients

Sigmoid

Tanh

ReLU

Swish

A. Reduces model parameters

B. Assigns weights to input tokens based on relevance

C. Prevents overfitting

D. Compresses the input sequence

Dropout

Gradient clipping

Batch normalization

Layer normalization

Both players make optimal decisions

Only one player uses a rational strategy

Players take random actions

The game is stochastic in nature

Hidden Markov Model (HMM)

Q-learning

Recurrent Neural Network (RNN)

Markov Decision Process (MDP)

Initial convolutional layers

All hidden layers

Output layer

Pooling layer

Features are independent given the class

The model uses Bayes' theorem

Features are conditionally dependent on each other

The prior probabilities are used