This directory contain multiple implementations of the CNN for handwritten image recognition that follow Zhang's design.
The code is located in cnn.apl, and the input files are located under input
directory. Note that if you are cloning the repo for the first time, you'd have
to manually download the MNIST data.
In order to run the code, you'd need to have an operational Dyalog APL interpreter. We have tried versions 16 and 17:
$ cat cnn.apl | dyalogWe explain many details in the following paper: https://dl.acm.org/doi/pdf/10.1145/3315454.3329960
We wrote another paper about SaC versions of the same CNN: https://dl.acm.org/doi/10.1145/3460944.3464312 The main motivation is to speed-up the building blocks that are required to define the network without loosing too much expressiveness.
The implementation of this CNN in SaC is here: https://github.com/SacBase/array-2021-supmaterial
There is an online video where the SaC paper is presented: https://www.pldi21.org/prerecorded_array.3.html