ntbuilder.h File Reference

Construction Functions.
http://tituxdev.github.io/NeuroTIC/src/CPU/ntbuilder.c. More...

#include "ntcore.h"
#include <stddef.h>

Include dependency graph for ntbuilder.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Macros
#define	NEWNET(network, neurons)   newnet( &network, neurons, (layer_t)sizeof( neurons )/sizeof( uint16_t ) )
	Convenience macro to create a new network.
#define	CREATE_NET_FEEDFORWARD(network, i, neurons)
	Convenience macro to create and build a feedforward network.

Functions
struct net_s *	newnet (net_s *net, uint16_t *neurons_per_layer, layer_t layers_size)
	Initializes a new neural network structure.
struct net_s *	buildnet (net_s *net)
	Builds the internal buffers and connections of a neural network.

Detailed Description

Construction Functions.
http://tituxdev.github.io/NeuroTIC/src/CPU/ntbuilder.c.

Provides memory allocation and pointer setup for neurons, input/output references, and inter-layer buffers.
Uses memtrack() to manage memory safely.

Author

Oscar Sotomayor

Date

2026

Definition in file ntbuilder.h.

Macro Definition Documentation

CREATE_NET_FEEDFORWARD

#define CREATE_NET_FEEDFORWARD	(		network,
			i,
			neurons
	)

Value:

    net_s network={ \
        .inputs= i, \
        .layers= sizeof( neurons )/sizeof( uint16_t ), \
    }; \
    buildnet( newfeedforward( NEWNET( network , neurons ) ) );

Convenience macro to create and build a feedforward network.

Combines newnet() and buildnet() into a single step for creating a feedforward network with the specified input size and layer configuration.

Parameters

network	Name of the network variable to create.
i	Number of external inputs to the network.
neurons	Array defining the number of neurons in each layer.

Returns

net_s Pointer to the fully constructed feedforward network.

NEWNET

#define NEWNET	(		network,
			neurons
	)		newnet( &network, neurons, (layer_t)sizeof( neurons )/sizeof( uint16_t ) )

Convenience macro to create a new network.

Simplifies calling newnet() by automatically calculating the number of layers from the neurons array size.

Parameters

network	Pointer to the network structure to initialize.
neurons	Array defining the number of neurons per layer.

Returns

net_s Pointer to the initialized network on success, or NULL on failure.

Function Documentation

buildnet()

struct net_s * buildnet

(

net_s *

net

)

Builds the internal buffers and connections of a neural network.

Parameters

net	Pointer to the network structure to build.

Returns

Pointer to the fully constructed network.

Allocates memory for inputs, outputs, and inter-layer buffers according to the bff_wiring configuration.
Sets neuron input pointers and weight arrays.
Buffer type explanations:

• 'M': Mixed buffer; each pointer is set based on src_type.
  • 'N': Points to another buffer in the network.
  • 'I': Points to the network input array.
  • 'O': Points to the network output array.
• 'N': Shared network buffer.
• 'I': Network input array.
• 'O': Network output array.

Validates wiring configurations and ensures all pointers are correctly set up for feedforward computation. Uses memtrack for all allocations.

References wiring_s::array_type, wiring_s::arrays, net_s::bff, neuron_s::bff_idx, neuron_s::in, net_s::in, neuron_s::inputs, net_s::inputs, net_s::layers, net_s::neurons, net_s::nn, neuron_s::out, net_s::out, wiring_s::size, wiring_s::src_index, wiring_s::src_layer, wiring_s::src_type, and net_s::wiring.

Referenced by loadnet().

newnet()

struct net_s * newnet	(	net_s *	net,
		uint16_t *	neurons_per_layer,
		layer_t	layers_size
	)

Initializes a new neural network structure.

Parameters

net	Pointer to the network structure to initialize.
neurons_per_layer	Array specifying the number of neurons in each layer.
layers_size	Number of layers (size of neurons_per_layer array).

Returns

Pointer to the initialized network on success, or NULL on failure.

Initializes a new neural network structure by allocating memory for neurons and layers, and setting initial pointers to NULL.
Validates input parameters and uses memtrack to register allocations for automatic cleanup.

References net_s::layers, and net_s::neurons.

Referenced by loadnet().