My PhD addresses a major problem in the computational neuroscience: The population coding. The population coding consists on analyzing a population of neurons since we believe that interactions (instantaneous and temporal) between spiking neurons make sense in the way that the brain processes and understands this information. In order to perform population coding, one needs tools that allow analyzing the dynamics of large networks. I developed a method that answers this first part of the work (see EnaS and this paper for more details). I used the maximal entropy principle and Montecarlo method to build the algorithmic analysis. From the analysis, I am interested in finding the statistical properties of the system (spiking network) dynamics. This is the first step of Analyzing. I am also interested in understanding the relation between the stimulation (natural movie) and the spiking data. The method that I developed to analyze spiking data is characterized by:
- Ability of large scale analysis.
- The CPU-time increases in a linear (not exponential) fashion with the network size.
- The dynamical model is general: one can personalize the model patterns as much as desired, e.g, Spatio-temporal (To allow introducing patterns with memory, e.g. not only firing rates and instantaneous pairwise interactions) and high-order interaction (not limited to pariwise, such as triplet and quadruplets interactions).
For this work, we have several collaborators with whom we communicate in order to exchange the expertise around the subject: