Statistical Inference for (in)tractable-likelihood problems from partially observed stochastic processes and (non-renewal) point processes

Description

In many signal-processing applications, it is of primary interest to decode/reconstruct the unobserved signal based on some partially observed information. Some examples are all type of recognition (e.g. automatic speech, face, gesture, handwriting), chemistry, genetics, genomics and neuroscience (ion channels modelling). From a statistical point of view, this corresponds to perform statistical inference of the underlying model parameters from fully/partially observed stochastic processes (e.g. discrete observations of one or more other coordinates) and (non-renewal) point processes (where each event is the epoch when a coordinate reaches/crosses a certain value, yielding the so-called first-passage-time problem). I will briefly present a couple of simple (but still realistic) examples (with application on cancer data and visual data) where the underlying likelihood function can be derived, leading to maximum likelihood estimation. Quite often though, due to the complexity of the models, the likelihood is unknown or intractable, requiring the investigation of new ad-hoc mathematical and statistical techniques to handle the so-called intractable-likelihood inference problems. Here I will focus on likelihood-free methods, and in particular on Approximate Bayesian Computation (ABC) method. I will illustrate it on two different examples arising from neuroscience, with data corresponding to either partially observed stochastic processes or non-renewal point processes.

Period	19 Sept 2018
Event title	Statistical Inference for (in)tractable-likelihood problems from partially observed stochastic processes and (non-renewal) point processes
Event type	Other
Location	ItalyShow on map