scirs2-neural 0.3.4

Neural network building blocks module for SciRS2 (scirs2-neural) - Minimal Version
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111

//! # Reinforcement Learning (`rl`)
//!
//! Self-contained deep RL implementations for both discrete and continuous action spaces.
//!
//! ## Sub-modules
//!
//! | Module | Description |
//! |--------|-------------|
//! | [`environments`] | [`Environment`] trait + `CartPole`, `PendulumEnv`, `GridWorld` |
//! | [`replay_buffer`] | Fixed-capacity circular replay buffer |
//! | [`policy`] | [`Policy`] trait, [`SimpleNetwork`], [`EpsilonGreedy`], [`BoltzmannPolicy`] |
//! | [`value`] | [`ValueNetwork`], [`QNetwork`], [`DuelingQNetwork`] |
//! | [`dqn`] | DQN / Double-DQN agent with ε-greedy exploration |
//! | [`actor_critic`] | A2C: synchronous Advantage Actor-Critic |
//! | [`ppo`] | Proximal Policy Optimisation (clipped surrogate + GAE) |
//! | [`sac`] | Soft Actor-Critic (maximum-entropy, off-policy) |

pub mod actor_critic;
pub mod dqn;
pub mod environments;
pub mod policy;
pub mod ppo;
pub mod replay_buffer;
pub mod sac;
pub mod value;

// ── Environments ─────────────────────────────────────────────────────────────

pub use environments::{
    ActionSpace,
    CartPole,
    ContinuousCartPole,
    Environment,
    GridCell,
    GridWorld,
    ObservationSpace,
    PendulumEnv,
};

// ── Policy ───────────────────────────────────────────────────────────────────

pub use policy::{
    BoltzmannPolicy,
    EpsilonGreedy,
    Policy,
    PolicyRng,
    SimpleNetwork,
    categorical_sample,
    softmax,
    softmax_temperature,
};

// ── Value networks ───────────────────────────────────────────────────────────

pub use value::{
    ActionValuePolicy,
    DuelingQNetwork,
    NetworkUpdate,
    QNetwork,
    SoftmaxValuePolicy,
    ValueNetwork,
};

// ── DQN ──────────────────────────────────────────────────────────────────────

pub use dqn::{
    DQNAgent,
    DQNConfig,
    DQNReplayBuffer,
    Experience,
};

// ── A2C ──────────────────────────────────────────────────────────────────────

pub use actor_critic::{
    A2CAgent,
    A2CConfig,
    A2CTrainInfo,
    ActorNetwork,
    CriticNetwork,
    run_episode,
};

// ── PPO ──────────────────────────────────────────────────────────────────────

pub use ppo::{
    ActorCritic,
    PPOConfig,
    PPOInfo,
    RolloutBuffer,
    PPO,
};

// ── SAC ──────────────────────────────────────────────────────────────────────

pub use sac::{
    Critic,
    SACConfig,
    SACInfo,
    StochasticActor,
    SAC,
};

// ── Replay buffer ────────────────────────────────────────────────────────────

pub use replay_buffer::{
    ReplayBuffer,
    Rng,
    Transition,
    XorShift64,
};