NeuroTIC

USER GUIDE
https://tituxdev.github.io/NeuroTIC/

Quick Start

---

Create a New Project

~/NeuroTIC$ make create 
 
Error: The workspace directory is not empty.
Do you want to remove all files in the workspace directory? (y/n): y
Workspace cleared.
 
Enter the new project name: test
Author name (default: whoami): whoami
Brief project description: testing
Available header packs:
basic
complete
Enter the header pack to use: complete
Project created successfully at  workspace/test.c

---

Write Your Code

// Filename: test.c
// Author: whoami
// Date: 2026-02-17
// Description: testing.
 
// Pro tip: use test.h to store global constants, macros, variables,
// or include additional headers. Keep your main clean.
 
#include "test.h"
#include "ntcomplete.h"
 
#define NETWORK_NAME test
#define OUTPUT_DIR "./test_net.ntic"
 
#define INPUTS 1
#define NEURONS_PER_LAYER 1 // , ... , ... , ... 
#define ACTIVATION_FUNCTIONS NTACT_SIGMOID // Watch ntactivation.h for available activation functions
 
#define TRAINING_SAMPLES 1
#define LEARNING_RATE 0.1
#define TOLERANCE 0.0
#define MAX_ATTEMPTS 1
 
int main( void ){
 
 // Your code here
 
 CREATE_NET_FEEDFORWARD( NETWORK_NAME , INPUTS , ( (uint16_t []){ NEURONS_PER_LAYER } ) ); 
 // Select activation functions for each neuron
 for( layer_t i= 0 ; i < NETWORK_NAME.layers ; i++ ) for( uint16_t j= 0 ; j < NETWORK_NAME.neurons[i] ; j++ )
     NETWORK_NAME.nn[i][j].fn= ACTIVATION_FUNCTIONS;
 // Initialize weights
 randnet( &NETWORK_NAME );
 
 // Prepare training data
 traindata_t data={
     .learning_rate= (precision_t)(LEARNING_RATE),
     .tolerance= (precision_t)(TOLERANCE),
     .max_attempts= MAX_ATTEMPTS,
     .samples= TRAINING_SAMPLES,
};
// Allocate training data
newtraindata( &data , &NETWORK_NAME );
// Fill training data with your samples and expected outputs
// for( sample_t i= 0 ; i < TRAINING_SAMPLES ; i++ ){
//     for( input_t j= 0 ; j < NETWORK_NAME.inputs ; j++ ) data.in[i][j]= ...; // Set input values
//     for( uint16_t j= 0 ; j < NETWORK_NAME.neurons[NETWORK_NAME.layers-1] ; j++ ) data.results[i][j]= ...; // Set expected output values
// }
 
 // Train the network
 backpropagation( &NETWORK_NAME , &data );
 
 // Select a test sample to evaluate the trained network
 size_t sample;
 printf( "Select test: " );
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wunused-result"
 scanf( "%zu" , &sample );
 #pragma GCC diagnostic pop
 // Print test
 printf( "SAMPLE INPUTS: " );
 for( input_t i= 0 ; i < NETWORK_NAME.inputs ; i++ ) printf( "%f " , data.in[sample][i] );
 printf( "\nEXPECTED OUTPUTS: " );
 for( uint16_t i= 0 ; i < NETWORK_NAME.neurons[NETWORK_NAME.layers-1] ; i++ ) printf( "%f " , data.results[sample][i] );
 
 //Connect network
 data_t inputs[NETWORK_NAME.inputs];
 for( input_t i= 0 ; i < NETWORK_NAME.inputs ; i++ ) NETWORK_NAME.in[i]= &inputs[i];
 // Set input values
 memcpy( inputs , data.in[sample] , NETWORK_NAME.inputs * sizeof( data_t ) );
 printf( "\n\nNETWORK INPUTS:  " );
 for( input_t i= 0 ; i < NETWORK_NAME.inputs ; i++ ) printf( "%f " , *NETWORK_NAME.in[i] );
 // Evaluate network
 feedforward( &NETWORK_NAME );
 // Print outputs
 printf( "\nNETWORK OUTPUTS: " );
 for( uint16_t i= 0 ; i < NETWORK_NAME.neurons[NETWORK_NAME.layers-1] ; i++ ) printf( "%f " , *NETWORK_NAME.out[i] );
 
 // Save the trained network to a file
 savenet( &NETWORK_NAME , OUTPUT_DIR );
 
 // Your code here
 
 putchar( '\n' );
 return 0;
}

API EXAMPLE

---

Compile

~/NeuroTIC$ make compile
make compile
Enter project location (e.g., workspace, examples): workspace
Enter project name (filename without extension): test
Enter platform (e.g., CPU): CPU
Compiling project in location: workspace
Project name: test
Platform: CPU

---

Run Your Project

~/NeuroTIC$ ./workspace/test
Select test: 0
SAMPLE INPUTS: 0.000000 
EXPECTED OUTPUTS: 0.000000
 
NETWORK INPUTS:  0.000000
NETWORK OUTPUTS: 0.496875