SethTrowbridge/nn
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

layer batch training ok

Browse Source
This commit is contained in:
TreetopFlyer 2021-07-29 10:13:21 -04:00
parent 23e71542f2
commit 1e644a70dc
3 changed files with 64 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
index.html
View File

@ -614,7 +614,7 @@ NN.Network.Create = function()
var i; var i;
obj.Layers = []; obj.Layers = [];
obj.LearningRate = 0.8; obj.LearningRate = 0.1;
obj.Error = []; obj.Error = [];
for(i=0; i<arguments.length-1; i++) for(i=0; i<arguments.length-1; i++)
@ -688,31 +688,33 @@ NN.Network.Stochastic = function(inNetwork, inTrainingSet, inIterations)
]; ];
let matrix2 = [ let matrix2 = [
[0.5793881115472015, 0.9732593374796092, 0.15207639877016987, -0.5356575655337803] [0.7098703863463034, 0.35485944251238033, 0.7642849892333241, 0.03046174288491077],
[-0.30655426258144347, 0.45509633551425077, -0.5013795222004322, -0.3421292736637427]
]; ];
let typeA = [ let input = [
[ 0.1, 0.05], [ 0.1, 0.05],
[ 0.0, -0.06] [ 0.0, -0.06],
[ 0.99, 0.85],
[ 1.2, 1.05]
]; ];
let typeB = [ let output = [
[ 0.99, 0.85], [1, 0],
[ 1.2, 1.05] [1, 0],
]; [0, 1],
let goals = [ [0, 1]
[1, 1, 0],
[0, 0, 1]
]; ];
var layer1 = NN.Layer.Create(1, 1); let nn1 = NN.Network.Create(2, 3, 2);
layer1.Forward.Matrix = matrix1; nn1.Layers[0].Forward.Matrix = matrix1;
nn1.Layers[1].Forward.Matrix = matrix2;
let stage1 = NN.Layer.Forward(layer1, typeA); let logLayers = inNN => inNN.Layers.forEach(L=>console.log(L.Forward.Matrix));
let stage1Error = NN.Layer.Error(layer1, goals);
let stage1Back = NN.Layer.Backward(layer1, stage1Error);
console.log("matrix before", layer1.Forward.Matrix); logLayers(nn1);
NN.Layer.Adjust(layer1, 0.1);
console.log("matrix after", layer1.Forward.Matrix); NN.Network.Batch(nn1, {Input:input, Output:output}, 100);
logLayers(nn1);
</script> </script>

71
nn.test.js
View File

@ -26,48 +26,62 @@ Deno.test("check.forward", ()=>
]; ];
let matrix2 = [ let matrix2 = [
[0.5793881115472015, 0.9732593374796092, 0.15207639877016987, -0.5356575655337803] [0.7098703863463034, 0.35485944251238033, 0.7642849892333241, 0.03046174288491077],
[-0.30655426258144347, 0.45509633551425077, -0.5013795222004322, -0.3421292736637427]
]; ];
let typeA = [ let input = [
[ 0.1, 0.05], [ 0.1, 0.05],
[ 0.0, -0.06] [ 0.0, -0.06],
[ 0.99, 0.85],
[ 1.2, 1.05]
]; ];
let typeB = [ let output = [
[ 0.99, 0.85], [1, 0],
[ 1.2, 1.05] [1, 0],
]; [0, 1],
let goals = [ [0, 1]
[1, 1, 0],
[0, 0, 1]
]; ];
let layers = [matrix1]; let layers = [matrix1, matrix2];
let stages = Forward(Methods.Mutate.Pad(typeA), layers); console.log("BEFORE", layers);
Backward(stages, layers, goals, 0.1); for(let i=0; i<100; i++)
{
let stages = Forward(Methods.Mutate.Pad(input), layers);
Backward(stages, layers, output, 0.1);
}
console.log("AFTER", layers);
}); });
/* /*
Deno.test("NN.Label", ()=> Deno.test("NN.Label", ()=>
{ {
Label(training, typeA, [1]); Label(training, typeA, [1, 0]);
Label(training, typeB, [0]); Label(training, typeB, [0, 1]);
stages.push(training[0]);
console.log(training);
assertEquals(training.length, 2, "input and output sets created"); assertEquals(training.length, 2, "input and output sets created");
assertEquals(training[0].length, training[1].length, "both sets have same length"); assertEquals(training[0].length, training[1].length, "both sets have same length");
assertEquals(training[0][0].length, 3, "padded input component"); assertEquals(training[0][0].length, 3, "padded input component");
assertEquals(training[1][0].length, 1, "unchanged label vector"); assertEquals(training[1][0].length, 2, "unchanged label vector");
});
Deno.test("NN.Forward", ()=>
{
let layer1 = M.Create.Box([-1, -1, -1], [1, 1, 1], 2);
let layer2 = M.Create.Box([-1, -1, -1], [1, 1, 1], 1);
layers.push(layer1);
layers.push(layer2);
console.log(training[0]);
stages = Forward(training[0], layers);
console.log(stages);
}); });
Deno.test("NN.Backward", ()=> Deno.test("NN.Backward", ()=>
{ {
let layer1 = M.Create.Box([-1, -1, -1], [1, 1, 1], 2); let copy1 = M.Create.Clone(layers[0]);
let layer2 = M.Create.Box([-1, -1, -1], [1, 1, 1], 1); let copy2 = M.Create.Clone(layers[1]);
let copy1 = M.Create.Clone(layer1);
let copy2 = M.Create.Clone(layer2);
layers.push(layer1);
layers.push(layer2);
for(let i=0; i<100; i++) for(let i=0; i<100; i++)
{ {
@ -79,11 +93,10 @@ Deno.test("NN.Backward", ()=>
}); });
Deno.test("NN.Forward", ()=> Deno.test("NN.Label", ()=>
{ {
console.log(Forward(stages, layers)); let stages = Forward(training[0], layers);
console.log(training[1]); console.log(stages[stages.length-1]);
}); });
*/ */

3
nn.ts
View File

@ -41,15 +41,12 @@ const Backward = (inStages:N, inLayers:N, inGoals:Cloud.M, inRate:number):N =>
let errorScaled:Cloud.M = M.Batch.Multiply(errorBack, M.Batch.Derivative(layerOutput)); let errorScaled:Cloud.M = M.Batch.Multiply(errorBack, M.Batch.Derivative(layerOutput));
errorBack = M.Batch.Affine(errorScaled, M.Create.Transpose(inLayers[i])); errorBack = M.Batch.Affine(errorScaled, M.Create.Transpose(inLayers[i]));
console.log("matrix before:", inLayers[i]);
errorScaled.forEach((inScaledError:Cloud.V, inIndex:number)=> { errorScaled.forEach((inScaledError:Cloud.V, inIndex:number)=> {
const deltas = M.Batch.Scale(M.Create.Outer(layerInput[inIndex], inScaledError), inRate); const deltas = M.Batch.Scale(M.Create.Outer(layerInput[inIndex], inScaledError), inRate);
inLayers[i] = M.Batch.Subtract(inLayers[i], deltas); inLayers[i] = M.Batch.Subtract(inLayers[i], deltas);
}); });
console.log("matrix after:", inLayers[i]);
} }
return inLayers; return inLayers;
}; };