TernaryLIF and refactoring of LIF to inherit from AbstractLIF

[srrisbud]

Signed-off-by: Risbud, Sumedh [email protected]

Issue Number: #150

Objective of pull request: To merge TernaryLIF behaviour along with related refactoring of the LIF Process

Pull request checklist

Your PR fulfills the following requirements:

• Issue created that explains the change and why it's needed
• Tests are part of the PR (for bug fixes / features)
• Docs reviewed and added / updated if needed (for bug fixes / features)
• PR conforms to Coding Conventions
• PR applies BSD 3-clause or LGPL2.1+ Licenses to all code files
• Lint (pyb) passes locally
• Build tests (pyb -E unit) or (python -m unittest) passes locally

Pull request type

Please check your PR type:

• Bugfix
• Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no api changes)
• Build related changes
• Documentation changes
• Other (please describe):

What is the current behavior?

• There is no TernaryLIF

What is the new behavior?

• There is a TernaryLIF, which spikes +1 or -1 depending upon which of the upper or lower thresholds was crossed.

Does this introduce a breaking change?

• Yes
• No

Supplemental information

• The development required refactoring the existing LIF Process to carve out an AbstractLIF Process.
• Both LIF and TernaryLIF inherit from AbstractLIF
• This organization can be followed for organizing lava/proc library in general

[bamsumit]

Also, the type hints are missing. This is not fully consistent across all the repo, but we would want to incorporate it to reduce technical debt later.

pip install flake8-annotations
cd src/lava/proc/lif/
flake8 --ignore=ANN101,W503,E501 .

[GaboFGuerra]

Thanks for the implementation! For now, I reviewed processes and processes tests. Will come back tomorrow to review models and models tests but posting now in case is useful.

[drager-intel]

Overall, I think this looks good.

In the next refactor, if we keep this Abstract parent class/inheritance style structure, I would take bias and bias_exp vars out of the AbstractLIF Process and make them part of a fixed-point mix-in for a child LIF class.

I would also allow for heterogeneous du, dv, v_th, etc within a LIF Process, in anticipation of the Loihi2 compiler.

But, we have some decisions to make about the organization about neuron Procs before then and I think these 2 items can wait a bit.

[GaboFGuerra]

Looks good! Thanks for addressing some of the comments.

PS: You might add new issues for the comments were the discussion was inconclusive so that we don't forget about it ;)