spires_read_weights

Read the reservoir’s recurrent weight matrix into a caller-provided buffer.

Signature

spires_status spires_read_weights(const spires_reservoir *r, double *buffer);

Parameters

Returns

spires_status — SPIRES_OK on success. On failure:

• SPIRES_ERR_INVALID_ARG — r or buffer is NULL.

Example

size_t N = spires_num_neurons(r);
double *W = malloc(N * N * sizeof(double));
if (!W) return 1;
 
if (spires_read_weights(r, W) != SPIRES_OK) {
    free(W);
    return 1;
}
 
/* Inspect a specific connection */
printf("W[0->1] = %.6f\n", W[1 * N + 0]);
 
free(W);

Notes

• No allocation. Unlike spires_copy_weights, this function writes into a buffer you provide. It is suitable for use in tight loops where you want to avoid repeated malloc/free calls.
• Buffer size. The caller must ensure buffer is at least num_neurons * num_neurons * sizeof(double) bytes. Use spires_num_neurons to query num_neurons before allocating.
• Row-major layout. buffer[i * N + j] is the connection weight from neuron j to neuron i. Zero entries indicate no connection (sparse matrix stored dense).

See Also

• spires_copy_weights — allocating variant that returns a fresh buffer.
• spires_coarse_grain — compress the reservoir using a weight threshold.
• spires_num_neurons — query the dimension of the weight matrix.