MesoNet: selecting a reproducible baseline

need ONE baseline i can follow all the way through. check normalization + readout before getting distracted by another repo

candidate task: MNIST recognition with STDP in Brian2. The notebook compares discussion-forum implementations, older paper code, coincidence-detection examples, BindsNET and snnTorch examples.

blocking issues: old Python/Brian dependencies; unclear normalization behavior in one Colab example; and examples built around specialized sensory modeling rather than the intended general learning experiment. The entry does not settle on a successful baseline yet.

selection criteria: can it run locally, are parameters documented, can weights be saved/read back, and can the implementation be matched to the paper? Keep source model behavior intact before changing equations.

• review question: which differences are porting changes versus intentional scientific changes? Record both before comparing the baseline against MesoNet. The later GenevieveF Diehl–Cook port becomes the concrete reference in the following entries.

• References saved in this journal section

• https://brian.discourse.group/t/brian2-implementation-of-mnist-training-with-stdp-an-attempt/257
• https://youtu.be/GTXTQ_sOxak
• https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2018.00024/full
• https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2015.00099/full
• https://www.sciencedirect.com/science/article/abs/pii/S0893608017302903

XeAe receptive-field mosaic

Planned model changes