A5019499516- Bayesian Modeling and Causal Inference
- Statistical Methods and Inference
- Advanced Causal Inference Techniques
- Machine Learning and Algorithms
- Bayesian Methods and Mixture Models
- Data Analysis with R
- Statistical Methods and Bayesian Inference
- Gene Regulatory Network Analysis
- Genetic diversity and population structure
- Domain Adaptation and Few-Shot Learning
- Optimal Experimental Design Methods
- Machine Learning in Healthcare
- RNA and protein synthesis mechanisms
- Anorectal Disease Treatments and Outcomes
- Pelvic floor disorders treatments
- Genetic Mapping and Diversity in Plants and Animals
- Transportation Planning and Optimization
- Colorectal Cancer Surgical Treatments
- Gene expression and cancer classification
- Cognitive Science and Mapping
- Vehicle Routing Optimization Methods
- Railway Systems and Energy Efficiency
- AI-based Problem Solving and Planning
- Genomics and Phylogenetic Studies
- demographic modeling and climate adaptation
Bern University of Applied Sciences
2016-2018
ETH Zurich
2010-2018
SIB Swiss Institute of Bioinformatics
2012-2014
University of Bern
2013-2014
Data61
2013
Inferring causal effects from observational and interventional data is a highly desirable but ambitious goal. Many of the computational statistical methods are plagued by fundamental identifiability issues, instability, unreliable performance, especially for large-scale systems with many measured variables. We present software provide some validation recently developed methodology based on an invariance principle, called invariant prediction (ICP). The ICP method quantifies confidence...
From observational data alone, a causal DAG is only identifiable up to Markov equivalence. Interventional generally improves identifiability; however, the gain of an intervention strongly depends on target, that is, intervened variables. We present active learning (that optimal experimental design) strategies calculating interventions for two different goals. The first one greedy approach using single-vertex maximizes number edges can be oriented after each intervention. second yields in...
Main approaches for learning Bayesian networks can be classified as constraint-based, score-based or hybrid methods. Although high-dimensional consistency results are available constraint-based methods like the PC algorithm, such have not been proved methods, and most of even shown to consistent in classical setting where number variables remains fixed sample size tends infinity. In this paper, we show that based on greedy equivalence search (GES) achieved with adaptive restrictions space...
Summary In many applications we have both observational and (randomized) interventional data. We propose a Gaussian likelihood framework for joint modelling of such different data types, based on global parameters consisting directed acyclic graph corresponding edge weights error variances. Thanks to the nature parameters, maximum estimation is reasonable with only one or few points per intervention. prove consistency Bayesian information criterion estimating Markov equivalence class graphs...
Tandem repeats (TRs) represent one of the most prevalent features genomic sequences. Due to their abundance and functional significance, a plethora detection tools has been devised over last two decades. Despite longstanding interest, TR is still not resolved. Our large-scale tests reveal that current detectors produce different, often nonoverlapping inferences, reflecting characteristics underlying algorithms rather than true distribution TRs in data. simulations show power detecting...
Biological systems are understood through iterations of modeling and experimentation. Not all experiments, however, equally valuable for predictive modeling. This study introduces an efficient method experimental design aimed at selecting dynamical models from data. Motivated by biological applications, the enables crucial experiments: it determines a highly informative selection measurement readouts time points.
Main approaches for learning Bayesian networks can be classified as constraint-based, score-based or hybrid methods. Although high-dimensional consistency results are available constraint-based methods like the PC algorithm, such have not been proved methods, and most of even shown to consistent in classical setting where number variables remains fixed sample size tends infinity. In this paper, we show that based on greedy equivalence search (GES) achieved with adaptive restrictions space...
The investigation of directed acyclic graphs (DAGs) encoding the same Markov property, that is conditional independence relations multivariate observational distributions, has a long tradition; many algorithms exist for model selection and structure learning in equivalence classes. In this paper, we extend notion DAGs to case interventional distributions arising from multiple intervention experiments. We show under reasonable assumptions on experiments, defines finer partitioning than hence...