StreamAlignedBatcher.h File Reference

C++ port of the Python StreamAlignedBatcher – builds batches with a fixed number of examples per class, grouped together, required for I_avg's per-class averaging (Eq. 7 of "Faster Predictive Coding Networks via Better Initialization"). More...

#include <vector>
#include <random>
#include <cstring>
#include <stdexcept>

Go to the source code of this file.

Classes
class	Deep::StreamAlignedBatcher
	Builds batches with a fixed, grouped number of examples per class – see the file-level note for why this exists and what it deliberately doesn't own. More...

Namespaces
namespace	Deep
	This file exists to provide both Tensor.h and IComputeBackend.h with device types without introducing circular dependencies.

Detailed Description

C++ port of the Python StreamAlignedBatcher – builds batches with a fixed number of examples per class, grouped together, required for I_avg's per-class averaging (Eq. 7 of "Faster Predictive Coding Networks via Better Initialization").

Kept deliberately SEPARATE from DiscriminativePCNetwork/SimplePCNetwork/ ConvPCNetwork – this is a standalone data-pipeline utility, not part of the core numerical library. Any consumer (C++, or via future bindings) uses it independently, feeding its output into whichever network class they're using via the existing ClampInput/ClampState-style API.

Does NOT own or copy the dataset – holds pointers + stride info only. Caller must keep X/Y/labels alive for the batcher's lifetime. GetBatch() gathers selected rows into caller-provided output buffers, so the same buffers can be reused across calls without repeated allocation.