https://github.com/oml-npa/fluxextra.jl

Additional layers and functions for Flux.jl.
https://github.com/oml-npa/fluxextra.jl

flux julia-language machine-learning

Last synced: 5 days ago
JSON representation

Additional layers and functions for Flux.jl.

• Host: GitHub
• URL: https://github.com/oml-npa/fluxextra.jl
• Owner: OML-NPA
• License: mit
• Created: 2021-01-30T15:38:01.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2021-08-05T11:00:24.000Z (about 3 years ago)
• Last Synced: 2024-09-29T04:21:15.580Z (5 days ago)
• Topics: flux, julia-language, machine-learning
• Language: Julia
• Homepage:
• Size: 212 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • Funding: .github/FUNDING.yml
  • License: LICENSE

Awesome Lists containing this project

README

        
[![Docs dev](https://img.shields.io/badge/docs-stable-blue.svg)](https://oml-npa.github.io/FluxExtra.jl/stable/)

[![CI](https://github.com/OML-NPA/FluxExtra.jl/actions/workflows/CI-main.yml/badge.svg)](https://github.com/OML-NPA/FluxExtra.jl/actions/workflows/CI-main.yml)

[![codecov](https://codecov.io/gh/OML-NPA/FluxExtra.jl/branch/main/graph/badge.svg?token=JROBFGEVQN)](https://codecov.io/gh/OML-NPA/FluxExtra.jl)

# FluxExtra

Additional layers and functions for the [Flux.jl](https://github.com/FluxML/Flux.jl) machine learning library.

## Layers

### Join

```

Join(dim::Int64)

Join(dim = dim::Int64)

```

Concatenates a tuple of arrays along a dimension `dim`. A convenient and type stable way of using `x -> cat(x..., dims = dim)`.

### Split

```

Split(outputs::Int64,dim::Int64)

Split(outputs::Int64, dim = dim::Int64)

```

Breaks an array into a number of arrays which is equal to `output` along a dimension `dim`. `dim` should we divisible by `outputs` without a remainder.

### Flatten

```

Flatten()

```

Flattens an array. A convenient way of using `x -> Flux.flatten(x)`.

### Addition

```

Addition()

```

A convenient way of using `x -> sum(x)`.

### Activation

```

Activation(f::Function)

```

A convenient way of using `x -> f(x)`.

### Identity

```

Identity()

```

Returns its input without changes. Should be used with a `Parallel` layer if one wants to have a branch that does not change its input.

## Normalizations

### [0,1]

```

norm_01!(data::Vector{T}) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Rescales each feature (last dimension) to be in the range [0,1]. Returns min and max values for each feature.

```

norm_01!(data::T,min_vals::T,max_vals::T) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Rescales each feature (last dimension) to be in the range [0,1].

### [-1,1]

```

norm_negpos1(data::Vector{T}) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Rescales each feature (last dimension) to be in the range [-1,1]. Returns min and max values for each feature.

```

norm_negpos1(data::T,min_vals::T,max_vals::T) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Rescales each feature (last dimension) to be in the range [-1,1].

### Zero center

```

norm_zerocenter!(data::Vector{T}) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Subtracts the mean of each feature (last dimension). Returns a mean value for each feature.

```

norm_zerocenter!(data::T,min_vals::T,max_vals::T) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Subtracts the mean of each feature (last dimension).

### Z-score

```

norm_zscore!(data::Vector{T}) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Subtracts the mean and divides by the standard deviation of each feature (last dimension). Returns mean and standard deviation values for each feature.

```

norm_zscore!(data::T,mean_vals::T,std_vals::T) where {F<:AbstractFloat,N,T<:Array{F,N}}

```

Subtracts the mean and divides by the standard deviation of each feature (last dimension).

## Other

Makes `Flux.Parallel` layer type stable when used with tuples.