Class RLEnvironmentBase

Base class for reinforcement learning environment

Hierarchy (View Summary)

RLEnvironmentBase

Index

Constructors

Properties

_epoch _rewardReviver

Accessors

epoch rewardReviver

Methods

clone reset sample_action setState state step test

Constructors

constructor

new RLEnvironmentBase(): RLEnvironmentBase
Returns RLEnvironmentBase

Properties

_epoch

_epoch: number

_rewardReviver

_rewardReviver: (arg0: RLStepResult) => number

Accessors

epoch

get epoch(): number
Epoch

Returns number
- Defined in lib/rl/base.d.ts:73

rewardReviver

set rewardReviver(reviver: (arg0: RLStepResult) => number): void
Reward reviver
Parameters
- reviver: (arg0: RLStepResult) => number
  Reward reviver function that returns modified reward value
Returns void
- Defined in lib/rl/base.d.ts:78

Methods

clone

clone(): RLEnvironmentBase
Returns cloned environment.

Returns RLEnvironmentBase
Cloned environment
- Defined in lib/rl/base.d.ts:84

reset

reset(): void
Reset environment.

Returns void
- Defined in lib/rl/base.d.ts:88

sample_action

sample_action(agent: any): any[]
Sample an action.
Parameters
- agent: any
  Agent
Returns any[]
Sampled action
- Defined in lib/rl/base.d.ts:121

setState

setState(state: any[], agent: any): void
Set new state.
Parameters
- state: any[]
  New state
- agent: any
  Agent
Returns void
- Defined in lib/rl/base.d.ts:100

state

state(agent: any): any[]
Returns current state.
Parameters
- agent: any
  Agent
Returns any[]
Current state
- Defined in lib/rl/base.d.ts:94

step

step(action: any[], agent: any): RLStepResult
Do action and returns new state.
Parameters
- action: any[]
  Actions to be performed by the agent
- agent: any
  Agent
Returns RLStepResult
state, reward, done
- Defined in lib/rl/base.d.ts:107

test

test(state: any[], action: any[], agent: any): RLStepResult
Do actioin without changing environment and returns new state.
Parameters
- state: any[]
  Environment state
- action: any[]
  Actions to be performed by the agent
- agent: any
  Agent
Returns RLStepResult
state, reward, done
- Defined in lib/rl/base.d.ts:115