vae

Documentation

Overview

Package vae implements Auto-Encoding Variational Bayes Algorithm

https://arxiv.org/pdf/1312.6114.pdf

Index

• Constants
• type Model
  • func (e Model) Feed(ds model.Dataset) model.FatModel

Constants

const DecoderCollection = "decoder"

DecoderCollection is the name of collection containing decoder model

const DefaultBatchSize = 32

default batch size for auto-encoders training

const EncoderCollection = "encoder"

EncoderCollection is the name of collection containing encoder model

const LatentCol = "Latent"

Latent is the default name of feature for the decoder

const RecoderCollection = "recoder"

RecoderCollection is the name of collection containing recoder model

Types

type Model

type Model struct {
	// size of hidden layer, half of input by default
	Hidden int
	// size of additional hidden layer, can be zero
	Hidden2 int
	// size of latent (encoder Output/decoder input) layer
	Latent int
	// latent layer tensor as output of encoder and input for decoder
	// vae.LatentCol by default
	Feature string
	// generative output for decoder
	// model.PredictedCol by default
	Predicted string
	// Mxnet Context
	// mx.CPU by default
	Context mx.Context
	// batch size
	// vae.DefaultBatchSize by default
	BatchSize int
	// random generator seed
	// random by default
	Seed int
	// optimizer config
	// nn.Adam{Lr:0.001} by default
	Optimizer nn.OptimizerConf
	// input width
	// normally it's calculated from features
	Width int
}

Model of the Variational Auto-Encoder

func (Model) Feed

func (e Model) Feed(ds model.Dataset) model.FatModel

Feed model with dataset