FiveTech Software tech support forums

Code: Select all | Expand

#define EPOCHS         100
#define LEARNING_RATE  0.1

function Main()

   LOCAL input_data := {0.5, 0.3, 0.2, 0.7, 0.9, 0.1, 0.4, 0.6, 0.8, 0.2, 0.5, 0.7}
   LOCAL true_output := {0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7}

   train_neural_network( input_data, true_output)

return nil 

FUNCTION sigmoid(x)
    RETURN 1 / (1 + EXP(-x))
   
FUNCTION sigmoid_derivative(x)
    RETURN x * (1 - x)

PROCEDURE initialize_parameters( weights_input_hidden, weights_hidden_output, bias_hidden, bias_output)
    FOR i := 1 TO 12
        FOR j := 1 TO 10
            weights_input_hidden[i, j] := hb_Random( 0, 1 ) / 1
        NEXT
    NEXT
   
    FOR i := 1 TO 10
        FOR j := 1 TO 7
            weights_hidden_output[i, j] := hb_Random( 0, 1 ) / 1
        NEXT
        bias_hidden[i] := 0.0
    NEXT

    FOR i := 1 TO 7
        bias_output[i] := 0.0
    NEXT

PROCEDURE forward_propagation( inputs, weights_input_hidden, weights_hidden_output, bias_hidden, bias_output, hidden_output, predicted_output)
    LOCAL hidden_input[10]

    FOR i := 1 TO 10
        hidden_input[i] := 0.0
        FOR j := 1 TO 12
            hidden_input[i] := hidden_input[i] + inputs[j] * weights_input_hidden[j, i]
        NEXT
        hidden_input[i] := hidden_input[i] + bias_hidden[i]
        hidden_output[i] := sigmoid(hidden_input[i])
    NEXT

    FOR i := 1 TO 7
        predicted_output[i] := 0.0
        FOR j := 1 TO 10
            predicted_output[i] := predicted_output[i] + hidden_output[j] * weights_hidden_output[j, i]
        NEXT
        predicted_output[i] := predicted_output[i] + bias_output[i]
        predicted_output[i] := sigmoid(predicted_output[i])
    NEXT

PROCEDURE calculate_loss( predicted_output, true_output )
    LOCAL loss := 0.0

    FOR i := 1 TO 7
        loss := loss + 0.5 * ((predicted_output[i] - true_output[i]) ^ 2)
    NEXT

    RETURN loss

PROCEDURE backward_propagation( inputs, true_output, hidden_output, predicted_output, weights_hidden_output, weights_input_hidden, bias_hidden, bias_output)
    LOCAL output_error[7], output_delta[7]
    LOCAL hidden_error, hidden_delta
    
    FOR i := 1 TO 7
        output_error[i] := true_output[i] - predicted_output[i]
        output_delta[i] := output_error[i] * sigmoid_derivative(predicted_output[i])
    NEXT

    FOR i := 1 TO 10
        hidden_error := 0.0
        FOR j := 1 TO 7
            hidden_error := hidden_error + output_delta[j] * weights_hidden_output[i, j]
        NEXT

        hidden_delta := hidden_error * sigmoid_derivative(hidden_output[i])

        FOR j := 1 TO 12
            weights_input_hidden[j, i] := weights_input_hidden[j, i] + inputs[j] * hidden_delta * LEARNING_RATE
        NEXT
        bias_hidden[i] := bias_hidden[i] + hidden_delta * LEARNING_RATE

        FOR j := 1 TO 7
            weights_hidden_output[i, j] := weights_hidden_output[i, j] + hidden_output[i] * output_delta[j] * LEARNING_RATE
        NEXT
        bias_output[i] := bias_output[i] + output_delta[i] * LEARNING_RATE
    NEXT

PROCEDURE train_neural_network( inputs, true_output )
    LOCAL weights_input_hidden[12, 10], weights_hidden_output[10, 7], bias_hidden[10], bias_output[7]
    LOCAL hidden_output[10], predicted_output[7], loss

    initialize_parameters(weights_input_hidden, weights_hidden_output, bias_hidden, bias_output)

    FOR epoch := 1 TO EPOCHS
        hidden_output = Array( 10 )
        predicted_output = Array( 7 )
       
        forward_propagation(inputs, weights_input_hidden, weights_hidden_output, bias_hidden, bias_output, @hidden_output, @predicted_output)

        loss := calculate_loss(predicted_output, true_output)

        backward_propagation(inputs, true_output, hidden_output, predicted_output, weights_hidden_output, weights_input_hidden, bias_hidden, bias_output)

        IF Mod(epoch, 100) == 0
            ? "Epoch", epoch, "Loss:", loss
        ENDIF
    NEXT

    ? "Training complete!"

RETURN