Monthly Archives: April 2016

Talk by Joris Mooij

You are all cordially invited to the AMLab colloquium on April 26 at 16:00 in C3.163, where Joris Mooij will give a talk titled “Automating Causal Discovery and Prediction“. Afterwards there are drinks and snacks!

Abstract: The discovery of causal relationships from experimental data and the construction of causal theories to describe phenomena are fundamental pillars of the scientific method. How to reason effectively with causal models, how to learn these from data, and how to obtain causal predictions has been traditionally considered to be outside of the realm of statistics. Therefore, most empirical scientists still perform these tasks informally, without the help of mathematical tools and algorithms. This traditional informal way of causal inference does not scale, and this is becoming a serious bottleneck in the analysis of the outcomes of large-scale experiments nowadays. In this talk I will describe formal causal reasoning methods and algorithms that can help to automate the process of scientific discovery from data.

Talk by Yash Satsangi

You are all cordially invited to the AMLab colloquium today, April 19 at 16:00 in C3.163, where Yash Satsangi will give a talk titled “Exploiting Submodular Value Functions for Scaling Up Active Perception“. Afterwards there are drinks and snacks!

Abstract: In active perception tasks, an agent aims to select sensory actions that reduce its uncertainty about one or more hidden variables. For example, a mobile robot takes sensory actions to efficiently navigate in a new environment. While partially observable Markov decision processes (POMDPs) provide a natural model for such problems, as the number of sensors available to the agent grows, the computational cost of POMDP planning grows exponentially with it, making POMDP planning infeasible with traditional methods. We propose greedy point-based value iteration (PBVI), a new POMDP planning method that uses greedy maximization to greatly improve scalability in the action space of an active perception POMDP. Furthermore, we show that, under certain conditions, including submodularity, the value function computed using greedy PBVI is guaranteed to have bounded error with respect to the optimal value function. We establish the conditions under which the value function of an active perception POMDP is guaranteed to be submodular. Finally, we present a detailed empirical analysis on a dataset collected from a multi-camera tracking system employed in a shopping mall. Our method achieves similar performance to existing methods but at a fraction of the computational cost leading to better scalability for solving active perception tasks. 

Talk by Philip Versteeg

You are all cordially invited to the AMLab colloquium today, April 12 at 16:00 in C3.163, where Philip Versteeg will give a talk titled “Causal inference and validation with micro-array data”. Afterwards there are drinks and snacks!

Abstract: In causality, one objective is to find cause and effect in a high dimensional setting. Often, identifiability of the true underlying graph is hard if not impossible when only observational data is available. A novel method, invariant causal prediction (ICP) is based on a relatively simple invariance principle. It is capable of exploiting a combination of observational and perturbed interventional experiments and provides confidence statements on the identified causal effects. In this talk, an application of ICP on high-dimensional micro-array experiments from the species Saccharomyces Cerevisiae is discussed. An attempt is made to validate the findings both internally and with external results.

Tom Claassen joins AMLab

Tom Claassen joined AMLab as a parttime postdoc (50%). Tom studied physics in Twente and worked for several years as a Systems Architect before doing his PhD on causal discovery  and logic at the Radboud University Nijmegen. Tom will work on causality as a team member of the VIDI project of Joris Mooij.

Talk by Martin Gullaksen

You are all cordially invited to a presentation on Friday, April 8th, from 16:00-17:00 in C1.112 by Martin Gullaksen on his master’s thesis entitled “Probabilistic Spatio-Temporal Inference in Early Embryonic Development. The case of Drosophila Melanogaster“. 

Abstract: Being able to infer gene regulatory networks from spatio-temporal expression
data is a major problem in biology. This thesis proposes a new dynamic Bayes
networks approach, which we benchmark by using the well researched gap gene
problem of the Drosophila melanogaster, with the capability of realistically
inferring gene regulatory networks and producing high quality simulations. The
thesis solves practical issues, currently associated with spatio-temporal gene
inference, such as computational time and parameter fragility, while obtaining
a similar gene regulatory network and matrix as our ground truth network. The
proposed modelling framework computes the gene regulatory network in 10-15
second on a modern laptop. Effectively removing the computational barrier of
the problem and allowing for future gene regulatory networks of greater gene
count to be processed. Besides producing a gene regulatory matrix our method
also produces high quality simulations of the gene activation levels of the gap
gene problem. In addition, unlike many competing problem formulations, the
proposed model is probabilistic in nature, hence allowing statistical inference
to be made. Finally, using Bayesian statistics, we perform robustness tests on
the topology of our proposed gene regulatory network and our regulatory
weights.