Help with Neural Network Visualization Bug In Processing

[winterchan]

@glv

``

Tried doing that, it seems as if the problem is related to the

NeuralNet.pde

class NeuralNet {
  int iNodes, hNodes, oNodes, hLayers;
  Matrix[] weights;

  NeuralNet(int input, int hidden, int output, int hiddenLayers) {
    iNodes = input;
    hNodes = hidden;
    oNodes = output;
    hLayers = hiddenLayers;

    weights = new Matrix[hLayers + 1];
    weights[0] = new Matrix(hNodes, iNodes + 1);
    for (int i = 1; i < hLayers; i++) {
      weights[i] = new Matrix(hNodes, hNodes + 1);
    }
    weights[weights.length - 1] = new Matrix(oNodes, hNodes + 1);

    for (Matrix w : weights) {
      w.randomize();
    }
  }

  void mutate(float mr) {
    for (Matrix w : weights) {
      w.mutate(mr);
    }
  }

  float[] output(float[] inputsArr) {
    Matrix inputs = weights[0].singleColumnMatrixFromArray(inputsArr);
    Matrix curr_bias = inputs.addBias();

    for (int i = 0; i < hLayers; i++) {
      Matrix hidden_ip = weights[i].dot(curr_bias);
      Matrix hidden_op = hidden_ip.activate();
      curr_bias = hidden_op.addBias();
    }

    Matrix output_ip = weights[weights.length - 1].dot(curr_bias);
    Matrix output = output_ip.activate();

    return output.toArray();
  }

  NeuralNet crossover(NeuralNet partner) {
    NeuralNet child = new NeuralNet(iNodes, hNodes, oNodes, hLayers);
    for (int i = 0; i < weights.length; i++) {
      child.weights[i] = weights[i].crossover(partner.weights[i]);
    }
    return child;
  }

  NeuralNet clone() {
    NeuralNet clone = new NeuralNet(iNodes, hNodes, oNodes, hLayers);
    for (int i = 0; i < weights.length; i++) {
      clone.weights[i] = weights[i].clone();
    }
    return clone;
  }

  void show(float x, float y, float w, float h, float[] vision, float[] decision) {
    float space = 5;
    float nSize = (h - (space * (iNodes - 2))) / iNodes;
    float nSpace = (w - (weights.length * nSize)) / weights.length;
    float hBuff = (h - (space * (hNodes - 1)) - (nSize * hNodes)) / 2;
    float oBuff = (h - (space * (oNodes - 1)) - (nSize * oNodes)) / 2;

    int lc = 0;  // Layer Count

    // DRAW NODES
    for (int i = 0; i < iNodes; i++) {  // DRAW INPUTS
      fill(vision[i] > 0 ? color(0, 255, 0) : color(255));
      stroke(0);
      ellipseMode(CORNER);
      ellipse(x, y + (i * (nSize + space)), nSize, nSize);
    }

    lc++;

    for (int a = 0; a < hLayers; a++) {
      for (int i = 0; i < hNodes; i++) {  // DRAW HIDDEN
        fill(color(255));
        stroke(0);
        ellipseMode(CORNER);
        ellipse(x + (lc * nSize) + (lc * nSpace), y + hBuff + (i * (nSize + space)), nSize, nSize);
      }
      lc++;
    }

    for (int i = 0; i < oNodes; i++) {  // DRAW OUTPUTS
      fill(decision[i] > 0 ? color(0, 255, 0) : color(255));
      stroke(0);
      ellipseMode(CORNER);
      ellipse(x + (lc * nSpace) + (lc * nSize), y + oBuff + (i * (nSize + space)), nSize, nSize);
    }

    lc = 1;

    // DRAW WEIGHTS
    for (int i = 0; i < weights[0].rows; i++) {  // INPUT TO HIDDEN
      for (int j = 0; j < weights[0].cols - 1; j++) {
        if (weights[0].matrix[i][j] < 0) {
          stroke(255, 0, 0);
        } else {
          stroke(0, 0, 255);
        }
        line(x + nSize, y + (nSize / 2) + (j * (space + nSize)), x + nSize + nSpace, y + hBuff + (nSize / 2) + (i * (space + nSize)));
      }
    }

    lc++;

    for (int a = 1; a < hLayers; a++) {
      for (int i = 0; i < weights[a].rows; i++) {  // HIDDEN TO HIDDEN
        for (int j = 0; j < weights[a].cols - 1; j++) {
          if (weights[a].matrix[i][j] < 0) {
            stroke(255, 0, 0);
          } else {
            stroke(0, 0, 255);
          }
          line(x + (lc * nSize) + ((lc - 1) * nSpace), y + hBuff + (nSize / 2) + (j * (space + nSize)), x + (lc * nSize) + (lc * nSpace), y + hBuff + (nSize / 2) + (i * (space + nSize)));
        }
      }
      lc++;
    }

    for (int i = 0; i < weights[weights.length - 1].rows; i++) {  // HIDDEN TO OUTPUT
      for (int j = 0; j < weights[weights.length - 1].cols - 1; j++) {
        if (weights[weights.length - 1].matrix[i][j] < 0) {
          stroke(255, 0, 0);
        } else {
          stroke(0, 0, 255);
        }
        line(x + (lc * nSize) + ((lc - 1) * nSpace), y + hBuff + (nSize / 2) + (j * (space + nSize)), x + (lc * nSize) + (lc * nSpace), y + oBuff + (nSize / 2) + (i * (space + nSize)));
      }
    }
  }
}

or the

NeuralNetworkWindow

class NeuralNetworkWindow extends PApplet {
  OthelloGame parent;
  AIPlayer aiPlayer;

  NeuralNetworkWindow(OthelloGame parent, AIPlayer aiPlayer) {
    this.parent = parent;
    this.aiPlayer = aiPlayer;
  }

  public void settings() {
    size(800, 800);
    println("NeuralNetworkWindow settings initialized");
  }

  public void setup() {
    background(255);
    println("NeuralNetworkWindow setup completed");
  }

  public void draw() {
    background(255);
    drawNeuralNetwork();
    println("Neural network visualization drawn in NeuralNetworkWindow");
  }

  void drawNeuralNetwork() {
    float x = 50;
    float y = 50;
    float w = width - 100;
    float h = height - 100;
    float[] vision = new float[64]; // Example input data
    float[] decision = aiPlayer.brain.output(vision); // Example output data
    aiPlayer.brain.show(x, y, w, h, vision, decision);
  }

  void updateVisualization() {
    redraw();
    println("Neural network visualization updated in NeuralNetworkWindow");
  }
}

i’m not sure what but when i use this code it causes the glitch, i can’d find exactly why