Skip to main content

mathverse_ai/
lib.rs

1//! AI/ML mathematical primitives: tensors, activations, losses, optimizers, attention, layers, models.
2//!
3//! # Features
4//!
5//! - N-dimensional tensor with broadcasting, reshaping, and element-wise math
6//! - Activation functions: ReLU, GELU, Swish, Mish, Softmax, Sigmoid, Tanh, and more
7//! - Loss functions: MSE, MAE, Huber, Cross-Entropy, Binary Cross-Entropy, KL Divergence, Hinge
8//! - Evaluation metrics: Accuracy, Precision, Recall, F1, Confusion Matrix, ROC AUC, R², Explained Variance
9//! - Stateful optimizers: SGD (with momentum), Adam, AdamW
10//! - Learning rate schedulers: Constant, Step Decay, Cosine Annealing, Linear Warmup
11//! - Attention math: QKV projection, scaled dot-product attention, multi-head attention
12//! - Positional encodings: sinusoidal (additive), rotary (RoPE)
13//! - Layer/Batch/RMS normalization
14
15#![warn(missing_docs)]
16#![forbid(unsafe_code)]
17#![warn(clippy::all, clippy::pedantic, clippy::nursery)]
18
19extern crate alloc;
20
21mod internal;
22
23pub mod tensor;
24pub mod activations;
25pub mod losses;
26pub mod metrics;
27pub mod optimizers;
28pub mod attention;
29pub mod autograd;
30pub mod layers;
31pub mod models;
32pub mod data;
33pub mod conv_transpose;
34pub mod recurrent;
35pub mod attention_adv;
36pub mod vision;
37pub mod generative;
38pub mod registry;
39
40pub use tensor::Tensor;
41pub use activations::{relu, sigmoid, softmax, gelu, swish, tanh as tanh_act};
42pub use losses::{mse, cross_entropy, binary_cross_entropy_with_logits};
43pub use metrics::{accuracy, f1, confusion_matrix};
44pub use optimizers::{Sgd, Adam, AdamW, LrScheduler, Schedule};
45pub use attention::{scaled_dot_product_attention, multi_head_attention, sinusoidal_encoding, apply_rope};
46pub use layers::{Linear, LayerNorm, BatchNorm, Dropout};
47pub use models::{Sequential, MLP, TransformerBlock, Activation};
48pub use data::{DataLoader, Batch, train_test_split};