____ _
| _ \ | |
| ' \ / \ | | | |) / _ \/ / | ' _ \| '_
\ / \ ' \ / ` |
| | | | / || | < / (| () | | | | | | | | | | | / | | | (| |
|| ||_|\,|| __|\_/|| || ||| || ||_|| ||_,|
Spike waveform classifier aimed at: 1- Removing noise during preprocessing for improved clustering 1.5- Output from classifier provides an additional high quality feature for clustering on 2- Recommending electrodes with neurons for user ease
== Note:
############################################################
############################################################ Aiming for a 2GB dataset, half and half for spikes and noise.
== Suggestion for updating dataset Add maximum diversity of both noise and spike clusters (i.e. if we have a total of 2000 neurons, and space for 100,000 waveforms, then every neuron should contribute 100,000/2000 waveforms if possible)
############################################################
############################################################ Aimed at being integrated into blech_clust/blech_process.py, prior to actually performing clustering
== Method of operation:
############################################################
############################################################
Dealing with waveforms of different lengths 1- Either have multiple models, 2- Or standardize waveform length ** Best waveform snapshot can be empirically determined
Getting more labelled data for "not spikes"
############################################################
############################################################ Experimentation and model + data tracking is done on Neptune.ai. These details are not currently available publicly. Dataset used here can be accessed at https://drive.google.com/drive/folders/1i1WPL7gt0ckvpuGVoZKfnu27bRRVUQEX?usp=sharing