Simple neural network in pure C

neuralnet.c

#include <stdio.h>
#include <math.h>
#include <stdlib.h>

// Sigmoid activation function
float sigmoid(float x) {
    return 1.0 / (1.0 + exp(-x));
}

// Prediction function with sigmoid activation
float predict(float temp, float humi, float w1, float w2, float bias) {
    float raw_output = w1 * temp + w2 * humi + bias;
    return sigmoid(raw_output);
}

// Training function with proper gradient descent
void train(float *w1, float *w2, float *bias) {
    float lr = 0.01;
    int epochs = 1000;  // Changed to int and increased epochs
    
    // Training data: [temperature, humidity, label]
    // 0 = not good day to go outside, 1 = good day to go outside
    float training_data[4][3] = {
        {73, 46, 0},  // High temp, low humidity -> not good
        {91, 85, 1},  // High temp, high humidity -> good
        {87, 41, 0},  // High temp, low humidity -> not good
        {71, 66, 1}   // Moderate temp, moderate humidity -> good
    };
    
    printf("Starting training...\n");
    
    for (int epoch = 0; epoch < epochs; epoch++) {
        float total_loss = 0.0;
        
        // Process each training sample
        for (int j = 0; j < 4; j++) {
            float temp = training_data[j][0];
            float humi = training_data[j][1];
            float expected = training_data[j][2];
            
            // Forward pass: make prediction
            float prediction = predict(temp, humi, *w1, *w2, *bias);
            
            // Calculate error (loss)
            float error = expected - prediction;
            total_loss += error * error;  // Squared error for monitoring
            
            // Backward pass: calculate gradients
            // For sigmoid: derivative = output * (1 - output)
            float sigmoid_derivative = prediction * (1 - prediction);
            
            // Update weights using gradient descent
            *w1 += lr * error * temp * sigmoid_derivative;
            *w2 += lr * error * humi * sigmoid_derivative;
            *bias += lr * error * sigmoid_derivative;
        }
        
        // Print progress every 100 epochs
        if ((epoch + 1) % 100 == 0) {
            printf("Epoch %d, Average Loss: %.6f\n", epoch + 1, total_loss / 4.0);
        }
    }
}

// Function to test the model with various inputs
void test_model(float w1, float w2, float bias) {
    printf("\n--- Model Testing ---\n");
    
    // Test cases
    float test_cases[][2] = {
        {25.0, 60.0},   // Mild temp, moderate humidity
        {35.0, 80.0},   // High temp, high humidity  
        {15.0, 40.0},   // Low temp, low humidity
        {30.0, 30.0},   // Moderate temp, low humidity
        {20.0, 90.0},   // Low temp, high humidity
        {40.0, 50.0}    // High temp, moderate humidity
    };
    
    int num_tests = sizeof(test_cases) / sizeof(test_cases[0]);
    
    for (int i = 0; i < num_tests; i++) {
        float temp = test_cases[i][0];
        float humi = test_cases[i][1];
        float prediction = predict(temp, humi, w1, w2, bias);
        
        printf("Temp: %.1f°C, Humidity: %.1f%% -> Prediction: %.3f ", 
               temp, humi, prediction);
        
        if (prediction > 0.5) {
            printf("(Good day to go outside!)\n");
        } else {
            printf("(Not a good day to go outside)\n");
        }
    }
}

int main() {
    // Initialize weights and bias
    float w1 = 0.1, w2 = 0.1, bias = 0.0;
    
    printf("=== Neural Network Weather Predictor ===\n");
    printf("Initial weights and bias: w1=%.4f, w2=%.4f, bias=%.4f\n\n", 
           w1, w2, bias);
    
    // Train the model
    train(&w1, &w2, &bias);
    
    printf("\nTraining completed!\n");
    printf("Final weights and bias: w1=%.6f, w2=%.6f, bias=%.6f\n", 
           w1, w2, bias);
    
    // Test the trained model
    test_model(w1, w2, bias);
    
    // Interactive testing
    printf("\n--- Interactive Testing ---\n");
    printf("Enter temperature and humidity (or -1 -1 to exit):\n");
    
    float temp, humi;
    while (1) {
        printf("Temperature (°C): ");
        scanf("%f", &temp);
        printf("Humidity (%%): ");
        scanf("%f", &humi);
        
        if (temp == -1 && humi == -1) {
            break;
        }
        
        float prediction = predict(temp, humi, w1, w2, bias);
        printf("Prediction: %.3f -> %s\n\n", 
               prediction,
               prediction > 0.5 ? "Good day to go outside!" : "Not a good day to go outside");
    }
    
    printf("Goodbye!\n");
    return 0;
}