Artificial neural network

Projects General Guidance
1 
#define NumberOf(arg) ((unsigned int) (sizeof (arg) / sizeof (arg [0]))) //calculates the amount of layers (in this case 4)

#include <NeuralNetwork.h>

unsigned int layers[] = {3, 9, 9, 1};
float *outputs; // 4th layer's outputs (in this case output)

//Default Inputs
float inputs[8][3] = {
  {0, 0, 0}, //0
  {0, 0, 1}, //1
  {0, 1, 0}, //1
  {0, 1, 1}, //0
  {1, 0, 0}, //1
  {1, 0, 1}, //0
  {1, 1, 0}, //0
  {1, 1, 1}  //1
};

// it is 1 for each layer [Pretrained Biases ]
float biases[] = {1, 1, 0.99308};

// it is 3*9 + 9*9 + 9*1  [Pretrained weights]
float weights[] = {
  -0.676266,  3.154561, -1.76689 ,
   1.589422, -2.340522,  1.447924,
   0.291685, -1.222407,  0.669717,
  -1.059862,  2.059782, -1.113708,
  -1.790229,  1.472432, -1.903783,
  -5.094713,  7.437615, -5.033135,
   2.341339,  3.370419,  2.185228,
  -3.887402,  1.453663, -3.861217,
  -1.555083,  2.943702, -0.472324,

  -1.171853, -0.45975 , -0.986132, -0.583541, -1.250889, -1.064349, -0.656225, -0.689616, -0.570443,
  -5.30186 ,  1.078257,  0.864669, -2.917707, -2.280059, -2.018297,  1.577451, -3.758011, -4.153339,
  -0.556209, -0.998336, -0.80149 , -0.232561, -1.087017, -1.286771, -1.034251, -0.05806 , -0.415967,
  -1.475901, -0.039556,  0.144446, -0.485774, -0.041879,  0.955343, -1.492304, -0.577319, -0.466558,
  -0.307791, -0.624868, -0.733248, -0.572921,  1.156592,  9.843138, -2.721857, -0.064086, -1.642469,
  -0.824234, -0.440457,  0.180901, -0.683897, -0.487519,  0.189743, -1.430297,  0.238511, -0.824287,
   0.251094, -3.009409, -1.58829 ,  0.590185,  0.597326, -5.243015,  2.710771,  2.596604,  0.969508,
  -1.344488,  2.618552,  0.642735, -0.947158, -0.286999,  3.797427, -2.443925, -0.833397, -1.654542,
  -0.138234, -0.931373, -0.183022, -0.493784, -0.784119, -0.275703, -2.113665,  0.761188, -0.810006,

  -0.049101,
  -6.781154,
   0.14872 ,
  -2.332737,
  -4.983434,
  -1.396086,
   10.86302,
  -5.551509,
  -1.648114

};

void setup()
{

  Serial.begin(9600);

  NeuralNetwork NN(layers, weights, biases, NumberOf(layers)); // Creating a NeuralNetwork with Pretrained Weights and Biases

  //Goes through all inputs
  for (int i = 0; i < NumberOf(inputs); i++)
  {
    outputs = NN.FeedForward(inputs[i]); // Feeds-Forward the inputs[i] to the first layer of the NN and Gets the output 
    Serial.println(outputs[0], 7); // prints the first 7 digits after the comma.
  }

  NN.print();

}

void loop() {
  
}

Hi anyone would like to give me some advice on How to change the code above, as my input would be unknown that comes in real-time when running the code.
so that the input is from Arduino analog pin A0 in array size of 10,
eg:{20,0,0,136,0,765,34,0,0,0}
each time feed into the input layer of the neural network.

i have done this coding for the array:

const unsigned int numReadings = 10;
unsigned int analogVals[numReadings];

void setup()
{
  Serial.begin(9600);
}

void loop()
{
  Serial.print("[");
  for (unsigned int i = 0; i < numReadings; i++)
  {
    analogVals[i] = analogRead(A0);
    delay(100); 
    Serial.print(analogVals[i]);
    if (i < numReadings - 1) Serial.print(" ");
  }
  Serial.println("]");
}
2

I am not really clear what you want to do, but you should test whether any Serial data is available before reading because otherwise you will get spurious values of -1 in the data

3

In the program, the inputs array is declared and initialized as 8x3. Your real time inputs have to be in multiples of 3 to match the dimensions of the NN input channels.

What is this network trained to recognize, and what do your analog input values represent?

4

Then there are many changes you have to make, and no point in posting the unmodified example code.

5

I am posting the example code, as it is a complete code for someone to understand how to modify on the input part to make it come from pin A0.
I am only asking for that part.
I couldn't test on my coding whether it has an error or not, and making discussion here, as I didn't manage to make the input feed from A0 yet.

6

for someone to understand how to modify on the input part to make it come from pin A0.

That will obviously depend on your new code, which will have a different input array.

Write the new neural network code, and if you get stuck getting the analog input data properly prepared, post again.

7

any of the arduino boards are obviously not going to be able to do any such AI calculations. or at least not to acheive anything practical with neural networks. arduino could be useful to send readings to another device for processing but why make arduino do more than that?