fix errors | add freshly generated examples

README.md

@@ -26,59 +26,97 @@ var character = String.fromCharCode(parseInt(output.join(''), 2));
 
 ### [MNIST [0-9]](http:https://yann.lecun.com/exdb/mnist/)
 
+```config.json```:
+
+```javascript
+{
+ "mnist": true,
+ "network": {
+ "hidden": 40,
+ "learning_rate": 0.1
+ }
+}
+```
+
 * **Neurons**
  * ```400``` input
  * ```40``` hidden
  * ```4``` output
-* **Learning rate**
- * ```0.1```
-* **Training set**
- * ```59999``` digits
-* **Testing set**
- * ```9999``` digits
-* **Measured success rate**
- * ```82.08820882088209%```
+* **Learning rate:** ```0.1```
+* **Training set:** ```60000``` digits
+* **Testing set:** ```10000``` digits
+* **Training time:** ```3 min 58 s 225 ms```
+* **Success rate:** ```78.06%```
 
 ### [a-z]
 
-* **Fonts**
- * sans-serif
- * serif
+```config.json```:
+
+```javascript
+{
+ "mnist": false,
+ "text": "abcdefghijklmnopqrstuvwxyz",
+ "fonts": [
+ "sans-serif",
+ "serif"
+ ],
+ "training_set": 2000,
+ "testing_set": 1000,
+ "image_size": 16,
+ "threshold": 400,
+ "network": {
+ "hidden": 40,
+ "learning_rate": 0.1
+ }
+}
+```
+
 * **Neurons**
- * ```400``` input
+ * ```256``` input
  * ```40``` hidden
  * ```8``` output
-* **Learning rate**
- * ```0.1```
+* **Learning rate:** ```0.1```
 * **Training set**
- * ```52000``` characters
- * **Sample**
- * ![abcdefghijklmnopqrstuvwxyz](https://raw.github.com/mateogianolio/mlp-character-recognition/master/examples/abcdefghijklmnopqrstuvwxyz.png)
-* **Testing set**
- * ```13000``` characters
-* **Measured success rate**
- * ```96.32307692307693%```
+ * **Size:** ```52000``` characters
+ * **Sample:** ![abcdefghijklmnopqrstuvwxyz](https://raw.github.com/mateogianolio/mlp-character-recognition/master/examples/abcdefghijklmnopqrstuvwxyz.png)
+* **Testing set:** ```26000``` characters
+* **Training time:** ```2 min 10 s 752 ms```
+* **Success rate:** ```91.77692307692308%```
 
 ### [0-9]
 
-* **Fonts**
- * sans-serif
- * serif
+```config.json```:
+
+```javascript
+{
+ "mnist": false,
+ "text": "0123456789",
+ "fonts": [
+ "sans-serif",
+ "serif"
+ ],
+ "training_set": 2000,
+ "testing_set": 1000,
+ "image_size": 16,
+ "threshold": 400,
+ "network": {
+ "hidden": 40,
+ "learning_rate": 0.1
+ }
+}
+```
+
 * **Neurons**
- * ```400``` input
+ * ```256``` input
  * ```40``` hidden
  * ```8``` output
-* **Learning rate**
- * ```0.1```
+* **Learning rate:** ```0.1```
 * **Training set**
- * **Size**
- * ```20000``` digits
- * **Sample**
- * ![0123456789](https://raw.github.com/mateogianolio/mlp-character-recognition/master/examples/0123456789.png)
-* **Testing set**
- * ```5000``` digits
-* **Measured success rate**
- * ```99.22%```
+ * **Size:** ```20000``` digits
+ * **Sample:** ![0123456789](https://raw.github.com/mateogianolio/mlp-character-recognition/master/examples/0123456789.png)
+* **Testing set:** ```10000``` digits
+* **Training time:** ```1 min 6 s 620 ms```
+* **Success rate:** ```99.22%```
 
 ## Configuration
 

config.json

@@ -1,14 +1,5 @@
 {
- "mnist": false,
- "text": "abcdefghijklmnopqrstuvwxyz",
- "fonts": [
- "sans-serif",
- "serif"
- ],
- "training_set": 2000,
- "testing_set": 500,
- "image_size": 20,
- "threshold": 400,
+ "mnist": true,
  "network": {
  "hidden": 40,
  "learning_rate": 0.1

main.js

@@ -49,7 +49,7 @@
  training = [],
  testing = [],
  settings = {
- size: chars,
+ size: config.text.length,
  height: config.image_size,
  text: config.text,
  fonts: config.fonts,
@@ -67,6 +67,9 @@
  return function(text, data) {
  var png = new PNG({ filterType: 4 });
  png.parse(data, parse(text, index));
+
+ if(index === 0)
+ fs.writeFileSync('./examples/' + text + '.png', data, 'base64');
  };
  }
 
@@ -149,7 +152,7 @@
  pixel = [],
  i, j, k, x, y;
 
- for(i = 0; i < chars; i++) {
+ for(i = 0; i < config.text.length; i++) {
  for(y = 0; y < data.height; y++) {
  for(x = i * config.image_size; x < (i * config.image_size + config.image_size); x++) {
  position = (data.width * y + x) << 2;
@@ -160,15 +163,15 @@
  chunk.push(
  pixel.reduce(function(previous, current) {
  return previous + current;
- }) > threshold ? 0 : 1
+ }) > config.threshold ? 0 : 1
  );
  pixel = [];
  }
  }
 
  chunk = tools.center(chunk);
 
- if(index < training_set) {
+ if(index < config.training_set) {
  training.push({
  input: chunk,
  output: ('00000000' + text.charCodeAt(i).toString(2)).substr(-8).split('').map(Number)

network.js

@@ -11,21 +11,30 @@
  log(' hidden:', network.layers.hidden[0].size, 'neurons.');
  log(' output:', network.layers.output.size, 'neurons.');
  log('learning rate:', rate, '\n');
- log('learning ...');
+ log('training with', length, 'inputs ...');
+
+ var start = process.hrtime();
 
  while(set.length) {
  object = set.pop();
 
  if(count % Math.round(length / 10) === 0)
- log('progress:', Math.round(100 * (count / length)) + '%');
+ log('progress:', Math.round(100 * (count / length)),'%');
 
  network.activate(object.input);
  network.propagate(rate, object.output);
 
  count++;
  }
 
- log('... done');
+ var elapsed = process.hrtime(start);
+ var time = {
+ minutes: Math.floor(elapsed[0] / 60),
+ seconds: elapsed[0] % 60,
+ milliseconds: Math.floor(elapsed[1] / 1000000)
+ };
+
+ log('... done', '(' + time.minutes, 'min', time.seconds, 's', time.milliseconds, 'ms)');
  log();
  };
 
@@ -38,7 +47,7 @@
  count = 0;
 
  // test on random inputs
- log('testing on', length, 'samples ...');
+ log('testing on', length, 'inputs ...');
  while(set.length) {
  object = set.pop();