Crate concision

Concision

Concision aims to be a complete machine learning library written in pure Rust.

Modules

consts

data

Data

error

func

Functional

gnn

Graph Neural Networks (GNN)

init

Initialization

kan

Kolmogorov-Arnold Networks (KAN)

linear

Linear Models

math

Mathematics

nn

ops

Operations

prelude

traits

transformer

Transformers

types

utils

Macros

builder

dimensional

AS

getters

toggle

Structs

Error

ModelShape

Enums

ErrorKind

PredictError

Propagate

Constants

D_MODEL

The default model size for any given model

EPSILON

The default epsilon value for floating point operations

Traits

Affine

Apply

A trait for applying a function to a type

ApplyGradient

ApplyOn

ArrayLike

AsComplex

Backward

Backward describes an object capable of backward propagation.

Compile

Config

A trait used to denote objects that may be used for configuring various items

Configuration

Configuration describes composite configuration objects; A configuration object is allowed to inherit from another configuration object

Conjugate

Context

Decrement

Decrement generally describes an object capable of decrementing itself;

DefaultLike

Entry

FillLike

FloorDiv

Forward

Forward describes an object capable of forward propagation.

ForwardIter

ForwardIter describes any iterators whose elements implement Forward. This trait is typically used in deep neural networks who need to forward propagate across a number of layers.

Increment

Init

InitInplace

Initialize

This trait provides the base methods required for initializing an ndarray with random values. Initialize is similar to RandomExt, however, it focuses on flexibility while implementing additional features geared towards machine-learning models; such as lecun_normal initialization.

InitializeExt

This trait extends the Initialize trait with methods for generating random arrays from various distributions.

IntoAxis

IntoComplex

Inverse

IsSquare

MaskFill

This trait is used to fill an array with a value based on a mask. The mask is a boolean array of the same shape as the array.

Matmul

Matpow

Module

A Module defines any object that may be used as a layer in a neural network. Config is a type that defines the configuration of the module; including any and all hyperparameters. Params is a type that defines the parameters of the module; typically references a Linear set of parameters { weights, bias }

NdLike

OfType

OnesLike

OrInsert

Predict

RoundTo

Sequence

A trait for sequential data structures; This trait is implemented for iterators that have a known length.

SequenceIter

Setup

Store

Toggle

Train

Transform

Unsqueeze

ZerosLike

Functions

concat_iter

Creates an n-dimensional array from an iterator of n dimensional arrays.

floor_div

genspace

hstack

inverse

linarr

linspace

rangespace

creates a matrix from the given shape filled with numerical elements [0, n) spaced evenly by 1

round_to

Round the given value to the given number of decimal places.

stack_iter

Creates a larger array from an iterator of smaller arrays.

tril

Returns the lower triangular portion of a matrix.

triu

Returns the upper triangular portion of a matrix.

vstack

Type Aliases

BoxError

A type alias for a boxed Error type that is Send, Sync, and 'static.

BoxResult

A type alias for a boxed Result which returns some object, T, and uses a BoxError as the error type.

NdResult

Result

A type alias for a Result with the crate’s Error type. Defaults to Result<(), Error>