Michael Beyeler
Michael Beyeler
- create file in the right subdirectory (e.g., `pulse2percept/examples/stimuli`) - filename must start with "plot_" (e.g. `plot_grating_stim.py`) - generate docs locally: `make doc` from root - verify docs locally: open...
Some models are named by Author-Year, others have a name (e.g., `ScoreboardModel`). Figure out a consistent naming scheme
We need to show an example of how to use the various optimizers (e.g., `GridSearchOptimizer`, #262) with the various predictors (e.g., `ShapePredictor`, #282)
- Use scikit-image's [`regionprops`](https://scikit-image.org/docs/stable/auto_examples/segmentation/plot_regionprops.html) to calculate essential spatial properties of percepts, such as area, centroids, and image moments. - Apply to every frame of the percept. - Cache the result...
The Nanduri model stores `all_r3` as (num pixels, num time steps) array, which in some cases can be huge. What is really needed is a thread-local buffer: https://stackoverflow.com/a/39923441
We need a way to calculate electric potentials for square, hex electrodes similar to Wiley & Webster for circular electrodes. There might be no analytical solution, but an approximation based...
We need a new doc entry that talks about how to use the model selection package in combination with the available models.
Add cross-validation for the optimizers in the `model_selection` subpackage. There are at least three options: * Implement a `crossval_predict` function (like scikit-learn) that takes in an optimizer. * Add new...
`model.find_threshold` should be re-implemented as a sklearn estimator with `fit` and `predict` methods. This way it is compatible with the optimizers from the model selection subpackage.