ARNI Biological Learning Working Group

1) Charlotte onboarded vision and audio MNIST dataloaders. She’s focusing on predictive coding for sequential tasks (e.g., audio data/moving MNIST).
2) Todd built a multi-modal predictive coding baselines showing that unsupervised representation learning is possible here.
3) Eivinas proposed a backprop based autoencoder and Hebbian based autoencoder (maybe Exponentiated Gradients?).
4) Nihal offered to onboard 3D MNIST and workin Hebbian learning rules.

There was also discussion of software engineering conventions (e.g., Github practices, configuration tooling, etc.).

Virtual Link: request @ [email protected]