cr.sparse.data.sparse_normal_blocks

cr.sparse.data.sparse_normal_blocks(key, D, K, B, S=1, cor=0.0, normalize_blocks=False)[source]

Generates representations where some blocks have normally distributed coefficients while others are zero.

Parameters

  • key – a PRNG key used as the random key.

  • D (int) – Dimension of the model space

  • K (int) – Number of nonzero blocks

  • B (int) – Length of each block

  • S (int) – Number of sparse model vectors (default 1)

  • cor (float) – Intra block correlation under AR-1 model

  • normalize_blocks (bool) – Normalize the nonzero coefficients in each block

Returns

A tuple consisting of (i) a vector/matrix of sparse model vectors (ii) active block numbers (iii) indices corresponding to the locations of nonzero entries

Notes: - Each active block of samples is normalized to unit norm.