Add Bearing Fault Detection (UORED) example akida1 - #27
dmclellandbc wants to merge 15 commits into
Conversation
sbruers-brainchip
left a comment
There was a problem hiding this comment.
Good one,
Mainly rewording of docstrings.
One question about the actual input shape worth checking: comment showed (1200, 35) but the code said (300, 140), so I assumed that was the correct one.
I feel like a lot of information is currently "hidden" in each file's docstring, so it's a bit hard to follow, but I am assuming this will be resolved by the soon to be notebooks.
sbruers-brainchip
left a comment
There was a problem hiding this comment.
Good one,
Mainly rewording of docstrings.
One question about the actual input shape worth checking: comment showed (1200, 35) but the code said (300, 140), so I assumed that was the correct one.
I feel like a lot of information is currently "hidden" in each file's docstring, so it's a bit hard to follow, but I am assuming this will be resolved by the soon to be notebooks.
Docstring suggestions, except README Co-authored-by: sbruers-brainchip <sbruers@brainchip.com>
This PR integrates a new Akida 1 model for Bearing Fault Detection (1D time series vibration data), using the UORED-VAFCLS dataset.
Two performance metrics are presented, to highlight a systematic methodological flaw in the evaluation of models on this (and related) datasets.
Two TODOs remain beyond this PR: