Sign inSign in
ConceptAI & ML

Policy gradient

A class of reinforcement learning algorithms that optimize the policy parameters directly by maximizing the expected cumulative reward.
as Fei-Fei Li, Justin Johnson & Serena Yeung defines itStanford UniversityThe wording closest to all 7 definitions
Watch it explained

Know first

Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.

Reinforcement learning3Gradient ascent2Gradient descent2Expectation1Expected reward1Expected reward objective1Log probabilities1Policy1

Opens up

Concepts that name this one as a prerequisite.

Lecturers

4

Each in their own terms, from the easiest level.

  • Sergey Levine

    UC Berkeley · Undergraduate

    An expression used to estimate the gradient of expected reward with respect to policy parameters, often used as a model-free reinforcement learning algorithm.
    • 0:08Policy Gradients I: Deriving the REINFORCE Estimator & Policy Objective
    • 16:38Variational Inference, Part 3
    • 0:24Model-Based Reinforcement Learning with Policies
    • 0:04Exploration of deep reinforcement learning architectures and algorithms
  • David Silver

    Google DeepMind · Undergraduate

    A vector of partial derivatives of the policy objective function with respect to the policy parameters, used to ascend to a local maximum.
    • 26:40Policy Gradient Methods
  • Jacob Andreas

    MIT · Undergraduate

    A reinforcement learning method used to optimize the language model parameters to maximize expected reward.
    • 1:13:59Language Models: Prompting, Chain-of-Thought, and Instruction-Tuning

Lectures that assume it

16

The first 12 are shown.

Connections

32

The lectures that define it or assume it, and the concepts next to it.

Concepts9

Show 3 more

Lectures23

Show 17 more

4 lecturers. Definitions are written by AI from each lecture’s transcript. A time with a play mark starts the lecture at the moment the lecturer says it.