bench

README

Bench LVis

bench_lvis is supposed to be an easy-to-understand, easy-to-run network that nevertheless has the same performance characteristics as the big LVis network. As compared to examples/bench_objrec, bench_lvis has much less code.

Currently includes:

• 4 layers, with a mix of random connections, local connections, dense connections

TODOs:

• Network doesn't converge. Not really a problem if we're just looking to test performance, but would be nice to get working.
• No GUI. Would be nice to make it easier to understand connectivity patterns.

Documentation

Overview

bench runs a benchmark model with 5 layers (3 hidden, Input, Output) all of the same size, for benchmarking different size networks. These are not particularly realistic models for actual applications (e.g., large models tend to have much more topographic patterns of connectivity and larger layers with fewer connections), but they are easy to run..

Index

• Variables
• func CenterPoolIdxs(ly emer.Layer, n int) []int
• func ConfigEpcLog(dt *etable.Table)
• func ConfigNet(b *testing.B, net *axon.Network, ...)
• func ConfigPats(pats *etable.Table, numPats int, inputShape [2]int, outputShape [2]int)
• func TrainNet(net *axon.Network, pats, epcLog *etable.Table, epcs int, verbose bool)

Variables

var ParamSets = params.Sets{
	{Name: "Base", Desc: "these are the best params", Sheets: params.Sheets{
		"Network": &params.Sheet{
			{Sel: "Prjn", Desc: "",
				Params: params.Params{
					"Prjn.Learn.LRate.Base": "0.1",
					"Prjn.SWt.Adapt.LRate":  "0.1",
					"Prjn.SWt.Init.SPct":    "0.5",
				}},
			{Sel: "Layer", Desc: "",
				Params: params.Params{
					"Layer.Inhib.ActAvg.Nominal": "0.08",
					"Layer.Inhib.Layer.Gi":       "1.05",
					"Layer.Act.Gbar.L":           "0.2",
				}},
			{Sel: "#Input", Desc: "",
				Params: params.Params{
					"Layer.Inhib.Layer.Gi": "0.9",
					"Layer.Act.Clamp.Ge":   "1.5",
				}},
			{Sel: "#Output", Desc: "",
				Params: params.Params{
					"Layer.Inhib.Layer.Gi": "0.70",
					"Layer.Act.Clamp.Ge":   "0.8",
				}},
			{Sel: ".BackPrjn", Desc: "top-down back-projections MUST have lower relative weight scale, otherwise network hallucinates",
				Params: params.Params{
					"Prjn.PrjnScale.Rel": "0.2",
				}},
		},
	}},
}

Functions

func CenterPoolIdxs

func CenterPoolIdxs(ly emer.Layer, n int) []int

CenterPoolIdxs returns the unit indexes for 2x2 center pools if sub-pools are present, then only first such subpool is used. TODO: Figure out what this is doing

func ConfigEpcLog

func ConfigEpcLog(dt *etable.Table)

func ConfigNet

func ConfigNet(b *testing.B, net *axon.Network, inputNeurDimPerPool, inputPools, outputDim,
	threadNeuron, threadSendSpike, threadSynCa int, verbose bool)

func ConfigPats

func ConfigPats(pats *etable.Table, numPats int, inputShape [2]int, outputShape [2]int)

func TrainNet

func TrainNet(net *axon.Network, pats, epcLog *etable.Table, epcs int, verbose bool)