mark6

package
v0.0.0-...-5f2bfed Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Jan 30, 2024 License: BSD-3-Clause Imports: 10 Imported by: 0

Documentation

Index

Constants

View Source
const (
	// FFWindow is the feedfoward window
	FFWindow = 64
	// ModelWindow is the window size
	ModelWindow = 128
	// ModelSamples is the number of samples
	ModelSamples = 256
	// Inputs is the number of inputs
	Inputs = 256
	// Outputs is the number of outputs
	Outputs = 32
)

Variables

This section is empty.

Functions

func Mark6

func Mark6()

Mark6 is the mark6 model

Types

type Rand

type Rand struct {
	Mean   float32
	StdDev float32
	Count  float32
}

Random is a random variable

type XNet

type XNet struct {
	Inputs  int
	Outputs int
	Rng     *rand.Rand
	E       XSet
	Q       XSet
	K       XSet
	V       XSet
}

Net is a net

func NewEmbeddingNet

func NewEmbeddingNet(seed int64, inputs, outputs int) XNet

NewEmbeddingNetNet makes a new network

func XNewNet

func XNewNet(seed int64, inputs, outputs int) XNet

XNewNet makes a new network

func (XNet) CalculateStatistics

func (n XNet) CalculateStatistics(systems []XSample) XSet

CalculateStatistics calculates the statistics of systems

func (*XNet) Fire

func (n *XNet) Fire(query, key, value Matrix) (float32, Matrix, Matrix, Matrix)

Fire runs the network

func (*XNet) FireEmbedding

func (n *XNet) FireEmbedding(value Matrix) (float32, Matrix, Matrix, Matrix)

Fire with embedding runs the network

type XSample

type XSample struct {
	Entropy float32
	Neurons []Matrix
	Outputs Matrix
	Out     Matrix
}

Sample is a sample of a random neural network

type XSet

type XSet [][]Random

Set is a set of statistics

func XNewStatistics

func XNewStatistics(inputs, outputs int) XSet

NewStatistics generates a new statistics model

func (XSet) Sample

func (s XSet) Sample(rng *rand.Rand, inputs, outputs int) []Matrix

Sample samples from the statistics

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL