Skip to content

Latest commit

 

History

History
28 lines (17 loc) · 1.52 KB

File metadata and controls

28 lines (17 loc) · 1.52 KB

Neural Predictive Model Tutorial

Here, we provide a brief overview of our neural predictive model.

For our competition, we have trained two baselines:

  • Our state-of-the-art CNN model
  • a simple linear-nonlinear (LN) model

Have a look our demo notebook as an introduction as to how our CNN model works.

We also show how to train a model on multiple datasets at once, compute its performance, and show the retinotopy of the V1 neurons by relating the learned RF position to the anatomical coordinates with V1: Generalization demo.

CPU vs GPU

We recommend to use a machine with a CUDA compatible GPU. On a CPU, the model training and evaluation is possible, too, but quite slow. To use CPU only, set device=cpu instead of device=cuda in the example notebooks.

Video Explanation

  • This Video goes into the details of how our model is built (prepared for ICLR 2021).

References