Module predictive_coding 

Predictive coding via XOR + popcount in packed bitstream domain.

Error = XOR(predicted, actual), magnitude = popcount(error) / L. No multiplications needed — maps to XOR gates + popcount tree on FPGA.

Functions

batch_prediction_error

Batch prediction error: n_neurons × n_inputs packed streams. Returns per-neuron surprise values.

predict_and_xor_ema

EMA predict-and-XOR loop for spike codec compression. Returns (error_matrix flattened row-major, correct_prediction_count).

predict_and_xor_lfsr

LFSR predict-and-XOR loop: bit-true with sc_bitstream_encoder.v. Returns (error_matrix flattened row-major, correct_prediction_count).

prediction_error_packed

Compute prediction error between two packed bitstreams via XOR + popcount. Returns error magnitude in [0, 1].

update_prediction_weights

STDP-like weight update: push prediction weight toward actual probability.

xor_and_recover_ema

EMA XOR-and-recover loop for spike codec decompression.

xor_and_recover_lfsr

LFSR XOR-and-recover loop for decompression.