Neural network

itmfar
02 Jan 2018, 19:43

I would glad if we can cooperate in this field

HTtrader
04 Jan 2018, 12:51

Hi itmfar,

I have an idea on how I am going to go about this, how much programming and knowledge of neural netoworks do you have? If you can understand this we might be able to collaborate on this. I plan to have 4 input neurons high low open and close with 3 output bull bear neutral and at this stage plan to back propogate the learning.

Regards,

Tony

itmfar
06 Feb 2018, 10:06

RE:

hungtonydang said:

Hi itmfar,

I have an idea on how I am going to go about this, how much programming and knowledge of neural netoworks do you have? If you can understand this we might be able to collaborate on this. I plan to have 4 input neurons high low open and close with 3 output bull bear neutral and at this stage plan to back propogate the learning.

Regards,

Tony

I used to program java and c# as a junior and I worked with weka api as a learning machine.

suradi
06 Feb 2018, 12:31

This should get you started with perceptrons :)

 private double perceptron1()
        {
            int[,] input = new int[,] 
            {
                {
                    1,
                    0
                },
                {
                    1,
                    1
                },
                {
                    0,
                    1
                },
                {
                    0,
                    0
                }
            };

            int[] outputs = 
            {
                0,
                1,
                0,
                0
            };


            Random r = new Random();

            double[] weights = 
            {
                r.NextDouble(),
                r.NextDouble(),
                r.NextDouble()
            };
            double learningRate = 500;
            double totalError = 1;
            int last = MarketSeries.Close.Count - 1;
            while (totalError > 0.2)
            {
                totalError = 0;
                for (int i = 0; i < 4; i++)
                {

                    int output = calculateOutput1(input[i, 0], input[i, 1], weights);
                    int error = outputs[i] - output;

                    weights[0] += learningRate * error * input[i, 0];
                    weights[1] += learningRate * error * input[i, 1];
                    weights[2] += learningRate * error * 1;

                    totalError += Math.Abs(error);

                }


            }
            //macd.Histogram
            double a1 = macd.Histogram[last - 1];
            double a2 = macd.Histogram[last - 1 - 1];
            double a3 = macd.Histogram[last - 1 - 1 * 2];
            double a4 = macd.Histogram[last - 1 - 1 * 3];
            double sum = (weights[0] * a1) + (weights[1] * a2) + (weights[0] * a3) + (weights[1] * a4);
            return (sum >= 0) ? 1 : 0;



        }

        private static int calculateOutput1(double a1, double a2, double[] weights)
        {

            double sum = a1 * weights[0] + a2 * weights[1] + 1 * weights[2];
            return (sum >= 0) ? 1 : 0;
        }