**Scientific Seminar: **

**Likelihood free Inference for time series dataset**

Post Doctoral Researcher

Department of Computer Science

Aalto University Finland

**5 July 2016****13:00 - 14:00****Room A14**

The time series dataset, with non i.i.d. observations, is one of the most frequent type of datasets seen in inferential problems where analytical forms of the likelihood is unavailable, e.g. simulator based models defined by ODE, PDE or SDE. Their abundance in Astrophysics (Simulating the formation of galaxies, stars, or planets), Evolutionary biology (Simulating evolution), Weather Prediction (Prediction of weather using numerical model), Ecology (Simulating species migration over time) and Health science (Simulating the spread of an infectious disease) raises a significant inferrential question for modeling their underlying complex stochastic models.

In this talk we illustrate the application of popular likelihood-free inference methods like Approximate Bayesian computation (ABC) and synthetic likelihood for time-series dataset. We also illustrate a generalized framework, by combining synthetic likelihood and ABC, for the estimation of posterior distribution in likelihood-free inference correspodingly through parametric and nonparametric approximation.