Predictive Analytics with TensorFlow 11.1: Reinforcement Learning

The video tutorial introduces reinforcement learning (RL) as a middle ground between supervised and unsupervised learning. It explains how RL involves an agent making decisions to maximize long-term rewards by balancing exploration and exploitation. Key concepts such as value function, policy, utility, and Q function are discussed, highlighting their roles in determining optimal actions. The tutorial also outlines the basic steps of RL algorithms: infer, do, and learn. Examples and diagrams illustrate these concepts, emphasizing their application in areas like robotics.

What is the primary goal of a reinforcement learning agent?

• To avoid any feedback from the environment

• To minimize the immediate reward

• To maximize the total reward in the long run

• To explore all possible actions equally

In reinforcement learning, what does the value function represent?

• The speed of learning

• The number of actions available

• The immediate reward of an action

• The long-term desirability of states

What is the term used for the strategy that an RL agent follows?

• Exploration

• Exploitation

• Reward

• Policy

What is the optimal policy in reinforcement learning?

• A policy that minimizes actions

• A policy that avoids exploration

• A policy that maximizes short-term rewards

• A policy that maximizes long-term rewards

What does the utility function in reinforcement learning predict?

• The immediate reward only

• The expected immediate reward plus future rewards

• The number of states

• The speed of learning

Which of the following is NOT a step in reinforcement learning algorithms?

• Learn

• Forget

• Do

• Infer

What is the purpose of the 'do' step in reinforcement learning?

• To select the best action based on current knowledge

• To execute an action and observe the outcome

• To update the policy

• To ignore the environment