The tuning_feedbackRF/ folder mostly consists of code to estimate the feedback-dependent contextual modulation of the movie-trained models, mostly replicating the in vivo experiments from Keller et al. 2020, presented in Figure 3.
Note that you have to estimate the aRFs of the model prior to making predictions to the centre-surround gratings. See tuning_retinotopy/README.md.
- predict_grating.py inference the movie-trained model(s) to predict responses to classical and inverse gratings.
- estimate_size_tuning_curve.py estimate the feedforward (ffRF) and feedback (fbRF) size tuning curves / receptive fields from the predicted responses to classical and inverse gratings.
- visualize_size_tuning_curve.py visualize the ffRF and fbRF from the predicted responses (Figure 3E).
- predict_onset.py inference the movie-trained model(s) to predict responses to classical and inverse gratings with 1000 repeats to estimate the response onset delay, as described in Keller et al. 2020.
- visualize_response_onset.py visualize response onset delay to classical and inverse gratings (Figure 3F).