gyx v0.1.25 Gyx.Core.Exp

This is data structure for representing an experience piece. This is what is returned to an agent when interacting (calling step/1) with the environment.

To consider

Usually, the experience pieces an agent gets from the environment, are stored in a replay buffer, so the learning method can access to certain experiences given a retrieval function.

These custom sampling techniques are responsability of the replay buffer module.

Use info key to store any additional metadata that could be useful for a replay buffer to consider when sampling. For example, a timestamp that could guarantee an atomic broadcasted replay buffer.

Link to this section Summary

Types

t()

Link to this section Types

t()

t() :: %Gyx.Core.Exp{
  action: number(),
  done: boolean(),
  info: map(),
  next_state: any(),
  reward: float(),
  state: any()
}

gyx

v0.1.25

gyx v0.1.25 Gyx.Core.Exp

To consider

Link to this section Summary

Types

Link to this section Types

t() t() :: %Gyx.Core.Exp{ action: number(), done: boolean(), info: map(), next_state: any(), reward: float(), state: any() }

t()

t() :: %Gyx.Core.Exp{ action: number(), done: boolean(), info: map(), next_state: any(), reward: float(), state: any() }