Methods, systems, and media for random semi-structured row-wise pruning in neural networks

A system comprising:

• a processor;
• a pseudo-random number generator;
• and
• a memory storing instructions which, when executed by the processor device, cause the system to:
• receive a pruning threshold indicating a desired compression ratio for a convolutional layer of a convolutional neural network;
• provide a seed value to a pseudo-random number generator;
• generate a row pruning mask identifying a plurality of kernel rows of a filter of the convolutional layer to be pruned, wherein generating the row pruning mask comprises:
• generating, using the pseudo-random number generator, a sequence of pseudo-random random numbers based on the seed value;
• and
• for each kernel row of a first filter of the convolutional layer, determining whether to identify the kernel row of the first filter as one of the plurality of kernel rows to be pruned based on a comparison the pseudo-random number to the pruning threshold;
• and
• generate a pruned filter in accordance with the generated row pruning mask and the first filter, the pruned filter comprising at least the kernel rows of the first filter not identified as kernel rows to be pruned.

A computer-readable medium having instructions tangibly stored thereon, wherein the instructions, when executed by a processing unit, causes the processing unit to:

• receive a pruning threshold indicating a desired compression ratio for a convolutional layer of a convolutional neural network;
• provide a seed value to a pseudo-random number generator;
• generate a row pruning mask identifying a plurality of kernel rows of a filter of the convolutional layer to be pruned, wherein generating the row pruning mask comprises:
• generating, using the pseudo-random number generator, a sequence of pseudo-random random numbers based on the seed value;
• and
• for each kernel row of a first filter of the convolutional layer, determining whether to identify the kernel row of the first filter as one of the plurality of kernel rows to be pruned based on a comparison the pseudo-random number to the pruning threshold;
• and
• generate a pruned filter in accordance with the generated row pruning mask and the first filter, the pruned filter comprising at least the kernel rows of the first filter not identified as kernel rows to be pruned.