Finish updating autoencoder to use loss function

src/autoencoder.test.ts

@@ -1,4 +1,5 @@
 import AutoencoderGPU from './autoencoder';
+import { INeuralNetworkTrainOptions } from './neural-network';
 
 const trainingData = [
   [0, 0, 0],
@@ -9,15 +10,22 @@ const trainingData = [
 
 const xornet = new AutoencoderGPU<number[], number[]>({
   inputSize: 3,
-  hiddenLayers: [5, 2, 5],
+  hiddenLayers: [4, 2, 4],
+  outputSize: 3,
 });
 
-const errorThresh = 0.011;
+const errorThresh = 0.0011;
 
-const result = xornet.train(trainingData, {
-  iterations: 100000,
+const trainOptions: Partial<INeuralNetworkTrainOptions> = {
   errorThresh,
-});
+  iterations: 250000,
+  learningRate: 0.1,
+  log: (details) => console.log(details),
+  // logPeriod: 500,
+  logPeriod: 500,
+};
+
+const result = xornet.train(trainingData, trainOptions);
 
 test('denoise a data sample', async () => {
   expect(result.error).toBeLessThanOrEqual(errorThresh);
@@ -58,7 +66,7 @@ test('test a data sample for anomalies', async () => {
   expect(result.error).toBeLessThanOrEqual(errorThresh);
 
   function includesAnomalies(...args: number[]) {
-    expect(xornet.likelyIncludesAnomalies(args)).toBe(false);
+    expect(xornet.likelyIncludesAnomalies(args, 0.5)).toBe(false);
   }
 
   includesAnomalies(0, 0, 0);

src/autoencoder.ts

@@ -28,8 +28,8 @@ function loss(
 ) {
   let error = expected - actual;
 
-  // if ( o ≈ i0 ) then return 10% of the loss value.
-  // Otherwise, return 1000% of the full loss value.
+  // if ( o ≈ i0 ) then return 3.125% of the loss value.
+  // Otherwise, return 3200% of the full loss value.
   // eslint-disable-next-line @typescript-eslint/ban-ts-comment
   // @ts-expect-error
   if (Math.round(actual) !== Math.round(inputs[this.thread.x])) error *= 32;
@@ -127,7 +127,10 @@ export class AutoencoderGPU<
    * @param {DecodedData} input
    * @returns {boolean}
    */
-  likelyIncludesAnomalies(input: DecodedData, anomalyThreshold = 0.2): boolean {
+  likelyIncludesAnomalies(
+    input: DecodedData,
+    anomalyThreshold: number
+  ): boolean {
     // Create the anomaly vector.
     const anomalies: number[] = [];
 
@@ -149,6 +152,8 @@ export class AutoencoderGPU<
     // Calculate the mean anomaly.
     const mean = sum / (input as number[]).length;
 
+    console.log(sum, mean, anomalyThreshold);
+
     // Return whether or not the mean anomaly rate is greater than the anomaly threshold.
     return mean > anomalyThreshold;
   }