Efficient coding of natural scenes improves neural system identification
Published in PLoS Computational Biology, 2023
This work explores how the efficient coding hypothesis can improve neural system identification models. A hybrid model predicting retinal neuron responses to noise stimuli was built by incorporating an autoencoder regularizer based on natural scene statistics, demonstrating improved prediction accuracy and more biologically plausible neural representations, particularly for direction-selective retinal neurons.
Recommended citation: Qiu, Y., Klindt, D. A., Szatko, K. P., Gonschorek, D., Hoefling, L., Schubert, T., Busse, L., Bethge, M., & Euler, T. (2023). "Efficient coding of natural scenes improves neural system identification." PLoS Computational Biology. 19(4), e1011037.
Download Paper