HopfieldNetwork-v2.1.0

What's Changed

• Add flag for profiling by @hmcalister in #310
• Add learn state data to nonintensive collection by @hmcalister in #314
• Add function to map binary vector to bipolar vector by @hmcalister in #315
• Update binary state energy functions by @hmcalister in #316
• Remove unneeded zeroing of vectors from learning rule loops by @hmcalister in #320
• Remove unneeded variables and calculations from thermal delta learnin… by @hmcalister in #321
• Add weight factor to thermal delta learning rule by @hmcalister in #322

Full Changelog: HopfieldNetwork-v2.0.1...HopfieldNetwork-v2.1.0

Hopfield Network

What's Changed

• Move learnStateData from intensive to non-intensive collection group by @hmcalister in #302
• Removes number of steps data field from unique states dataframe by @hmcalister in #304
• Add distance measure by @hmcalister in #308

Full Changelog: HopfieldNetwork-v2.0.0...HopfieldNetwork-v2.0.1

Hopfield Network

What's Changed

• Update delta learning rule to CORRECTLY apply delta learning by @hmcalister in #234
• Pull reused method of creating constant vector to function by @hmcalister in #236
• Distances by domain by @hmcalister in #239
• Update hopfield network to use new domain manager distance functions by @hmcalister in #241
• Update delta rule to avoid ugly double subtraction by @hmcalister in #243
• Learning rate implementation by @hmcalister in #246
• Learning rate summary by @hmcalister in #248
• Bias vector by @hmcalister in #250
• Force zero bias flag by @hmcalister in #252
• Save bias vector to a file by @hmcalister in #254
• Refactor learning rule and method by @hmcalister in #256
• Refactor domain distance metrics by @hmcalister in #258
• Add bias to energy by @hmcalister in #261
• Rename symmetric weight matrix flag by @hmcalister in #263
• Add force zero bias to summary by @hmcalister in #265
• Add rand matrix init flag by @hmcalister in #267
• Single step update delta learning by @hmcalister in #269
• Update energy calculations by @hmcalister in #271
• Add forceZeroBias field to network struct by @hmcalister in #273
• Fix updateBias calculation to include previous terms by @hmcalister in #275
• Fix energy calculations by @hmcalister in #286
• Update hebbian learning rule by @hmcalister in #287
• Update delta learning rule by @hmcalister in #288
• Add thermal delta learning rule by @hmcalister in #289
• Add progress bar by @hmcalister in #290
• Use single rows for state updates by @hmcalister in #291
• Moves saving of target state vectors to sensible location by @hmcalister in #292
• Add progress bar to saving relaxation results by @hmcalister in #294
• Correct delta update factor to 0.5 by @hmcalister in #296
• Fix network update by @hmcalister in #298
• Add more concurrent writers to data handlers by @hmcalister in #300

Full Changelog: HopfieldNetwork-v1.6.0...HopfieldNetwork-v2.0.0

Hopfield Network

What's Changed

• Learning method refactor by @hmcalister in #203
• Rename learningRuleMap to avoid pluar by @hmcalister in #205
• Iterative batch learning method by @hmcalister in #207
• Domain enum by @hmcalister in #209
• Add stringer for domain enum by @hmcalister in #211
• Add learning mapping functions to allow for abstraction of learning rules by @hmcalister in #213
• Update activation function to use new domain to select correct function by @hmcalister in #215
• Add domain to hopfield network builder by @hmcalister in #217
• Add domain to hopfield network by @hmcalister in #219
• Update calls to activation functions by @hmcalister in #221
• Update state generator to use new domain arg by @hmcalister in #223
• Update summary to also track the hopfield network domain by @hmcalister in #225
• Add domain arg to main.go and flags by @hmcalister in #228
• Fix binary learning mapping by @hmcalister in #230
• State manager by @hmcalister in #232

Full Changelog: HopfieldNetwork-v1.5.0...HopfieldNetwork-v1.6.0

Hopfield Network

What's Changed

• Removes artefacts of presentation ordering from delta learning rule by @hmcalister in #171
• Add stable field to learn state data by @hmcalister in #173
• Make the HopfieldNetwork use the LearnStateData struct explictly by @hmcalister in #175
• Implement stable field in learn state data by @hmcalister in #177
• Shift collection of learn state data to after learning step by @hmcalister in #179
• Expand network summary data by @hmcalister in #181
• Update range expression in delta learning rule by @hmcalister in #183
• Handler cleanup function by @hmcalister in #185
• Add unique state handler by @hmcalister in #187
• Explicitly set the state vector to always have the first unit be posi… by @hmcalister in #189
• Apply activation function to learned states before learning by @hmcalister in #191
• Move unique states collection out of intensive data collection method by @hmcalister in #194
• Add documenation for uniqueStates.pq by @hmcalister in #195
• Revert bipolar activation symmetry break by @hmcalister in #197
• Track state uniqueness using energy profile rather than state by @hmcalister in #199
• Correct title casing on relaxationHistory documentation by @hmcalister in #201

Full Changelog: HopfieldNetwork-v1.4.0...HopfieldNetwork-v1.5.0

Hopfield Network

What's Changed

• Remove saving of probe states binary file entirely by @hmcalister in #141
• Update the data processing logging to be consistent with other log st… by @hmcalister in #143
• Add epoch to logging during network learning by @hmcalister in #145
• Fix the logging message for epoch, formatting requires additional space by @hmcalister in #147
• Learn state data by @hmcalister in #149
• Rename learnState data file for consistency by @hmcalister in #151
• Remove presented index by @hmcalister in #154
• Rename history index to step index by @hmcalister in #157
• Rename history index to step index by @hmcalister in #158
• Update readme with data file documentation by @hmcalister in #159
• Move learn state data to hopfield network by @hmcalister in #162
• Flag allow intensive data collection by @hmcalister in #163
• Update comments in main by @hmcalister in #165
• Update main to actually remove the data directory before simulation by @hmcalister in #167
• Fix bug where energy profile was not correctly copied to history array by @hmcalister in #169

Full Changelog: HopfieldNetwork-v1.3.0...HopfieldNetwork-v1.4.0

Hopfield Network

What's Changed

• Add final state to relaxation result by @hmcalister in #125
• Fix number of states when loading from file by @hmcalister in #127
• Update target state probe to include state by @hmcalister in #129
• Removes the code of conduct (not needed for this project) by @hmcalister in #131
• Add logging messages by @hmcalister in #133
• Update learningRule flag description to remove additional space by @hmcalister in #135
• Create go.yml by @hmcalister in #137
• Add workflow badge to readme by @hmcalister in #139

Full Changelog: HopfieldNetwork-v1.2.0...HopfieldNetwork-v1.3.0

Hopfield Network

What's Changed

• Update data collector to rename targetstateprobe by @hmcalister in #96
• Update noise application to include factor of maximum network weight by @hmcalister in #97
• Gonum io by @hmcalister in #98
• Remove state aggregate collector by @hmcalister in #100
• Data collector documentation refactor by @hmcalister in #101
• Update energy function documentation by @hmcalister in #102
• Update learning rule documentation by @hmcalister in #103
• Update noise application docstring by @hmcalister in #105
• Improve LearningRule Docstrings by @hmcalister in #106
• Improve States Docstrings by @hmcalister in #107
• Improve HopfieldNetwork Docstrings by @hmcalister in #108
• Improve HopfieldUtils Docstrings by @hmcalister in #109
• Docstring update by @hmcalister in #110
• Gonum matrix io refactor by @hmcalister in #112
• Create CODE_OF_CONDUCT.md by @hmcalister in #113
• Update project to rename TestStates to ProbeStates by @hmcalister in #116
• Update main.go with additional comments by @hmcalister in #118
• Binary file io by @hmcalister in #119
• Rename command line flags for probeStates and targetStates to be more… by @hmcalister in #122
• Add logging message to end of file by @hmcalister in #123

Full Changelog: HopfieldNetwork-v1.1.0...HopfieldNetwork-v1.2.0

Hopfield Network

What's Changed

• Network data summary by @hmcalister in #80
• Remove trial by @hmcalister in #82
• Remove trial index from data collector structs by @hmcalister in #84
• Additional noise applications by @hmcalister in #86
• Asymmetric weight matrix by @hmcalister in #89
• Fix bug where array length is off by one for states that do not relax by @hmcalister in #90
• Target state data collection by @hmcalister in #93

Full Changelog: HopfieldNetwork-v1.0.0...HopfieldNetwork-v1.1.0

Hopfield Network

A first (and well overdue) release of this project. In future, releases should be made weekly perhaps?

What's Changed

• Learningrulestart by @hmcalister in #4
• Update random matrix init to use very small values, not significant ones by @hmcalister in #6
• Add chunking function by @hmcalister in #10
• Multiunit update by @hmcalister in #11
• Logging by @hmcalister in #12
• Make trials a command line argument by @hmcalister in #15
• Add number of test states as command line argument by @hmcalister in #16
• State NAN bugfix by @hmcalister in #22
• Update chunking to correctly chunk rather than have many repeated chunk values by @hmcalister in #23
• Relaxation return value by @hmcalister in #24
• Parquet data implementation by @hmcalister in #26
• Make threads a command line arg by @hmcalister in #28
• Add randomly generated seeds to distributions, using a random gen seeded from the system time as a base by @hmcalister in #29
• Data writer refactor by @hmcalister in #33
• Data handler refactor by @hmcalister in #35
• Remove dynamicstruct from go mod and sum by @hmcalister in #37
• MAJOR Overhaul to data collection system by @hmcalister in #38
• Symmetric units by @hmcalister in #40
• Code profile by @hmcalister in #43
• Add logging to stdout and keyboard interrupt handling by @hmcalister in #44
• Update activation functions to optimize for speed by @hmcalister in #45
• Update various scripts by @hmcalister in #46
• Move activation function problem from activation function (expensive)… by @hmcalister in #47
• Update scripts and data collection by @hmcalister in #48
• Change activation timing to correct position by @hmcalister in #52
• Revert continuous activations by @hmcalister in #56
• Concurrent delta by @hmcalister in #58
• Different activation fn by @hmcalister in #59
• Remove domain by @hmcalister in #61
• Data file truncation by @hmcalister in #63
• Data writer fix by @hmcalister in #65
• Data handler refactor by @hmcalister in #68
• Refactor data collector to use abstract addHandler method by @hmcalister in #69
• Relaxation history collection by @hmcalister in #70
• Networking logging by @hmcalister in #72
• Flag cleanup by @hmcalister in #74
• Update flags to have nicer layout in code by @hmcalister in #75
• Noisy delta by @hmcalister in #76
• Change keyboard interupt handling to profiler by @hmcalister in #78

New Contributors

• @hmcalister made their first contribution in #4

Full Changelog: https://github.com/hmcalister/Hopfield-Network-Go/commits/HopfieldNetwork-v1.0.0