MLP API

Multi-layer perceptron construction, evaluation, and parameters. More...

Data Structures
struct	mg_mlp
	Multi-layer perceptron model. More...

Functions
bool	mg_mlp_init (mg_graph *g, mg_mlp *m, size_t n_in, const size_t *sizes, size_t n_sizes)
	Initialize an MLP model.
bool	mg_mlp_call (mg_graph *g, mg_mlp *m, mg_value **x, mg_value **out)
	Evaluate an MLP model.
void	mg_mlp_free (mg_mlp *m)
	Free memory owned by an MLP model.
size_t	mg_mlp_param_count (const mg_mlp *m)
	Get the number of trainable parameters in an MLP model.
void	mg_mlp_params (const mg_mlp *m, mg_value **out)
	Copy an MLP model's parameters into a flat array.

Detailed Description

Multi-layer perceptron construction, evaluation, and parameters.

Function Documentation

mg_mlp_call()

bool mg_mlp_call	(	mg_graph *	g,
		mg_mlp *	m,
		mg_value **	x,
		mg_value **	out
	)

Evaluate an MLP model.

Calls each layer in the model sequentially.

Parameters

g	Graph that owns any intermediate and output values.
m	Model to evaluate.
x	Array of input values.
out	Output array with enough entries for the final layer.

Returns

true on success, false on invalid input or allocation failure.

mg_mlp_free()

void mg_mlp_free

(

mg_mlp *

m

)

Free memory owned by an MLP model.

Parameters

m	Model to free.

mg_mlp_init()

bool mg_mlp_init	(	mg_graph *	g,
		mg_mlp *	m,
		size_t	n_in,
		const size_t *	sizes,
		size_t	n_sizes
	)

Initialize an MLP model.

Parameters

g	Graph that owns the model's values.
m	Model to initialize.
n_in	Number of input features.
sizes	Array containing each layer's output size.
n_sizes	Number of entries in sizes.

Returns

true on success, false on allocation failure.

mg_mlp_param_count()

size_t mg_mlp_param_count

(

const mg_mlp *

m

)

Get the number of trainable parameters in an MLP model.

Parameters

m	Model to inspect.

Returns

Total number of parameters across all layers.

mg_mlp_params()

void mg_mlp_params	(	const mg_mlp *	m,
		mg_value **	out
	)

Copy an MLP model's parameters into a flat array.

Parameters

m	Model to inspect.
out	Output array with at least mg_mlp_param_count(m) entries.