neural coding & computation group
Generalized Integrate-and-Fire model - matlab code
description: performs simulation and maximum-likelihood estimation of a stochastic, leaky, integrate-and-fire model with linear receptive field and post-spike current (closely related to the "Spike-Responses Model", Jolivet et al 2003). This model is more general and more flexible than the traditional IF model, capable of exhibiting a rich array of dynamical behaviors and providing a highly accurate description of stimulus-evoked spike responses and their statistics. Surprisingly, the likelihood function is log-concave, meaning that ML estimation recovers the globally optimal model parameters (consisting of a stimulus filter, post-spike current waveform, leak conductance, leak reversal potential, and noise amplitude).
version 1: code_gif_v1.tgz (more info: README.txt)
sample figures (pdf)
mathematical details: notes on calculating likelihood (Fokker-Planck methods), excerpted from my Ph.D. thesis ( pdf)
- Pillow, JW, Paninski, L., Uzzell, VJ, Simoncelli, EP, and Chichilnisky, EJ. (2005). Prediction and Decoding of Retinal Ganglion Cell Responses with a Probabilistic Spiking Model. Journal of Neuroscience 25:11003-11013. (pdf)
- Paninski, L, Pillow, JW, and Simoncelli, EP. (2004). Maximum likelihood estimation of a stochastic integrate-and-fire neural encoding model. Neural Computation, 16:2533-2561. (pdf)