16. Quiz

It ensures that returns are always positive.

It simplifies the calculation of the reward for each individual trade.

It reduces the effect of outliers in the reward function.

It allows the cumulative return from a series of trades to be calculated accurately.

The rewards should be set at equal values to simplify comparisons.

Adjustments should be done iteratively to observe their impact on the agent's behavior.

Buy and hold rewards should be set as high as possible to ensure maximum profitability.

Adjustments should be made only once during the training phase.

It ensures that the agent only exploits the DQN predictions.

It randomly selects actions with a fixed probability.

It balances exploration and exploitation by choosing a random action with probability epsilon.

It updates the Q-values directly based on the selected actions.

To manage the exploration-exploitation balance

To generate random numbers for the epsilon-greedy strategy

To select actions based on the target Q-table

To update the DQN model by learning from past experiences

A reward for the chosen action

A new state after the action is performed

An action based on the current state

The Q-values for all possible actions

To increase the probability of selecting random actions as training progresses

To reduce the probability of selecting random actions as the model learns

To ensure the model always exploits the best action without exploration

To reset the exploration parameter to its initial value periodically

To ensure that random elements of training, such as actions and experiences, are consistent across runs

To increase the randomness of training and improve exploration

To minimize the size of the experience replay mini-batch

To accelerate the convergence of the training process

By comparing the training loss with the loss on a validation set

By checking the model's accuracy on the training dataset

By assessing the randomness of the trades performed

By evaluating the profitability of trades on a test dataset