This repository has programs which are used to test storing a large finite-state automaton using both a high dimensional superposition vector and a sparse distributed memory and comparing the performance of both methods when they are using the same amount of storage space.
The programs were used to generate figures for a paper submitted to Frontiers in Neuroscience, section Neuromorphic Engineering. Title is:
On separating long- and short-term memories in hyperdimensional computing
Verson 2, submitted Oct 3, 2022:
Script "plot_eedb.py" generates all graphs shown in the paper. To run it type:
python plot_eedb.py
The program reads data stored in a sqlite3 database named
empirical_error_noroll_v21.db.
That database was generated by running script build_eedb.py.
The build_eedb.py script imports scripts used to compute the
empirical error.
The predicted sizes and errors are in file mdie.py. The
contents of that file (the dimensions and errors) were
generated by running script find_sizes2.py. It
imports scripts that compute the predicted error for the
different memory variations.
For version 1 of the paper, submitted in February 2022:
The programs were used to generate the figures for a paper submitted in February 2022, titled:
On separating long- and short-term memories in hyperdimensional computing
Instructions for generating the figures
are in the README file in the figures_for_paper directory.