Class DQNAgent

constructor

• new DQNAgent(
      env: RLEnvironmentBase,
      resolution: number,
      layers: LayerObject[],
      optimizer: string,
  ): DQNAgent

  Parameters

  • env: RLEnvironmentBase

    Environment
  • resolution: number

    Resolution of actions
  • layers: LayerObject[]

    Network layers
  • optimizer: string

    Optimizer of the network

  Returns DQNAgent

_env

_net

method

• set method(value: "DQN" | "DDQN"): void

  DQN Method

  Parameters

  • value: "DQN" | "DDQN"

    New method name

  Returns void

get_action

• get_action(state: any[], greedy_rate?: number): any[]

  Returns a action.

  Parameters

  • state: any[]

    Current states
  • Optionalgreedy_rate: number

    Greedy rate

  Returns any[]

  Action

get_score

• get_score(): number[][][]

  Returns a score.

  Returns number[][][]

  Score values

terminate

• terminate(): void

  Returns void

update

• update(
      action: any[],
      state: any[],
      next_state: any[],
      reward: number,
      done: boolean,
      learning_rate: number,
      batch: number,
  ): number

  Update model.

  Parameters

  • action: any[]

    Action
  • state: any[]

    Current states
  • next_state: any[]

    Next states
  • reward: number

    Reward
  • done: boolean

    Done epoch or not
  • learning_rate: number

    Learning rate
  • batch: number

    Batch size

  Returns number

  Loss value