Example data and code for:

Jan Clemens, Mala Murthy (2021) _Quadratic and adaptive computations yield an efficient representation of song in Drosophila auditory receptor neurons, preprint

Requires matlab. See github.com/janclemenslab/glm_utils for code that uses python’s scikit-learn to fit a quadratic filter.

Directory structure:

• src/: Source code. Contains code from Park and Pillow (2011).
• dat/: Data files containing responses for different types of acoustic stimuli used in the paper.
• res/: Data files containing the results of model fitting.
• fig/: Figures with expected results.
• fit_model.m: Loads stimuli-response data from dat/, fits the quadratic&adaptive model, and saves results to res/.
• plot_predictions.m, plot_eigendecomposition.m: Loads model from ‘res/‘ and plots model predictions and the eigenvalue decomposition of the quadratic filter. See ‘fig/` for expected results.

Usage:

• run fit_models.m twice, with the filename in lines 4-5 set to dat/noise_20160311_8.mat and dat/step_20140625_1.mat, respectively. This will fit the model for two variants of the noise stimulus and save the results in res/.
• To plot the results, run plot_predictions.m and plot_eigendecomposition.m. This will load and plot the results from res/.