Antti Hyttinen



Docent / Adjunct Professor in Computer Science (University of Helsinki)
Ph. D. (Computer Science, University of Helsinki)
M. Sc. (Tampere University (of Technology))

email:



News and Highlights

  • From December 2022 working in industry, so no review requests!
  • Area chairing at UAI 2022.
  • K. Rantanen successfully defended his thesis, co-supervised by Prof. M. Järvisalo and A. Hyttinen, on 8.12.2021. Check out the news item and thesis.
  • I will be lecturing the Probabilistic Graphical Models -course in University of Helsinki in the spring 2022.
  • Slides from machine learning coffee seminar presentation on 26.10.2020 w. Jussi Viinikka can be found here.
  • Our paper 'Towards Scalable Bayesian Learning of Causal DAGs' was accepted to NeurIPS 2020!
  • We received the best paper award at the 23rd International Conference on Discovery Science! September 2020.
  • Journal paper 'Do-search - a tool for causal inference and study design with multiple data sources' accepted to Epidemiology, July 2020.
  • Past 500 citations in Google Scholar, June 2020.
  • Received the title of docent (i.e. adjunct professor) in Computer Science, University of Helsinki, 21.2.2020.


Short Bio

Dr. Antti Hyttinen obtained his Master’s degree Information Technology in Tampere University (of Technology) and his PhD in Computer Science in 2013 from the University of Helsinki. Dr. Hyttinen did research at California Institute of Technology during 2014. Dr. Hyttinen has also obtained the competitive personal post-doc funding from the Academy of Finland in 2016. Dr. Hyttinen received the title of docent (adjunct professor) in February 2020. Currently, Dr. Hyttinen works as a university researcher in the Sums of Products research group, collaborating with several research groups at the department.

Dr. Hyttinen's research focuses on causal inference and probabilistic graphical models within AI, machine learning and computational data analysis. Dr. Hyttinen has developed theory and structure learning algorithms for several different types of probabilistic graphical models. Many of these are causal models that allow also for latent confounding and feedback. These methods build on many different kinds of techniques such as Branch and Bound, MIP, MaxSAT, ASP, and dynamic programming. In addition, Dr. Hyttinen has developed methods for causal effect estimation, experiment selection and combining different types of data sources for the aforementioned tasks.



Student Supervision

  • Advisor for the Ph.D. research of Jelin Leslin (supervised by Prof. Martin Andraud, Aalto U.), 2021-.
  • Supervisor of Ph.D. thesis and research of Kari Rantanen (w. Prof. Matti Järvisalo), 2017-2021.
  • Supervising internship and M.Sc. thesis of Vitoria Barin-Pacela (w. Aapo Hyvärinen), 2020-2021.
  • Instructed 5 B.Sc. thesis in Computer Science, spring 2020. Topics included CNNs, LSTMs, ICA, and recommender systems.
  • Supervised (w. Michael Mathioudakis) summer internship of Riku Laine ((awarded) paper published in DS, see publication list).
  • Supervised (w. Matti Järvisalo) M. Sc. thesis of Kari Rantanen, 2017 (paper published in NIPS, see publication list).
  • Instructed 5 B.Sc. thesis in Computer Science, 2017. Topics included CNNs, decision trees, and argumentation frameworks.



Teaching

  • Lecturer of Probabilistic Graphical Models, spring 2022.
  • Lecturer of Probabilistic Graphical Models, spring 2021.
  • Instructor for the B.Sc. thesis course, spring 2020.
  • Lecturer of Probabilistic Graphical Models, spring 2019.
  • Lecturer of Probabilistic Graphical Models, spring 2018.
  • Instructor for the B.Sc. thesis course, fall 2017.
  • Lecturer of Probabilistic Models, spring 2016.
  • Course Assistant for Introduction to Machine Learning 2010.
  • Course Assistant for Causal Analysis 2008.


Software

CSI-do-search. Identification of causal effects using context-specific independence relations. See more details in the NeurIPS 2019 paper. Implementation by Santtu Tikka. 2018-2020. (R-package)
MAGSL. Exact search for learning maximal ancestral graphs with linear Gaussian BIC score. Includes local score calculation and pruning software. Implementation by Kari Rantanen, 2021. (code)
Do-search. Identification of causal effects from arbitrary observational and experimental probability distributions via do-calculus using a search-based algorithm. Allows for selection bias, transportability, missing data and arbitrary combinations of these. Implementation by Santtu Tikka. 2018-2020. (R-package)
Gadget and Beeps. Estimation of linear causal effects when the true graph is not known (Beeps+Gadget). Also state-of-the-art MCMC sampling for DAGs, with which one can estimate posterior probabilities of any structural features (edges, paths) of a DAG model. Implementation by Jussi Viinikka. 2020. (Code)
Imputation example.
bcause. The bcause program implements a practical exact problem-oriented branch-and-bound algorithm for discovering causal structure with latent confounding and feedback. Independence constraints are weighted according to their reliability and exact optimal solution with respect to these is found. Now supports interventional data and the sigma-separation criterion. Implementation by Kari Rantanen. 2020-2022. (Code)
Imputation example.
BIDA. Bayesian estimation of linear causal effects when the true graph is unknown. Includes also a method for estimating posterior probabilities of the presence of ancestral paths in a DAG, exactly via dynamic programming. This latter works with any local Bayesian network score. Implementation by Johan Pensar and Topi Talvitie. 2020. (Code)
Decompostion of a DAG.
Constraint-based Causal Discovery with ASP. An ASP-based approach for causal discovery in the presensence of latent confounding variables and cycles. The UAI2014 paper ASP and R code. (Code)
Imputation example.
Counterfactual-based imputation (CFBI). An approach for evaluation automatic decision makers from case data. More details in the DS2020 paper. Built on Stan. Implementation by Riku Laine. (Code)
Imputation example.
lsmarkov. Stochastic local search for chordal Markov networks. Implementation by Kari Rantanen. 2020. (Code)
bbmarkov. An exact branch-and-bound algorithm for discovering chordal Markov networks. Implementation by Kari Rantanen. 2017. (Code)
Chordal Markov Network.
Time series subsampling. Recovers causal time series structure when it is is observed only at coarser time points than the true time scale. See papers published in PGM2016 and IJAR2017 for details. (Code)
Subsampling time series example.
SAT-discoverer. A SAT-based approach for causal discovery in the presence of latent confounding variables and cycles. Only works with oracle data, so it can inspect the equivalence class structure. See Constraint-based Causal Discovery with ASP and bcause for further developments handling sample data. Implementation by Patrik Hoyer. 2013. (Code)
Imputation example.

Only most recent and/or most important software included here, see the publication list for further code packages.

Publications

See DBLP for most of the Bibtex entries.

2022

J. Leslin, A. Hyttinen, K. Periasamy, L. Yao, M. Trapp, M. Andraud
A Hardware Perspective to Evaluating Probabilistic Circuits
11th International Conference on Probabilistic Graphical Models, PGM 2022.
(Article)

A. Hyttinen, V. Barin-Pacela, A. Hyvärinen
Binary Independent Component Analysis: A Non-stationarity-based Approach
38th Conference on Uncertainty in Artificial Intelligence, UAI 2022.
(Article) (Poster) (Code)

2021

S. Tikka, A. Hyttinen, J. Karvanen
Causal Effect Identification from Multiple Incomplete Data Sources: A General Search-based Approach
Journal of Statistical Software, Articles, Volume 99, Issue 5, 2021.
(Article) (Preprint) (Code)

K. Rantanen, A. Hyttinen, M. Järvisalo
Maximal Ancestral Graph Structure Learning via Exact Search
Proceedings of the 37th Conference on Uncertainty in Artificial Intelligence, UAI 2021.
(Article) (Code)

J. Karvanen, S. Tikka, A. Hyttinen
Do-search - a tool for causal inference and study design with multiple data sources
Epidemiology (journal), Volume 32, Issue 1, p. 111-119, 2021.
(Article) (Preprint) (Code)

2020

J. Viinikka, A. Hyttinen, J. Pensar, M. Koivisto
Towards Scalable Bayesian Learning of Causal DAGs
Advances in Neural Information Processing Systems 33, NeurIPS 2020.
(Article) (Preprint) (Supplement) (Code)

R. Laine, A. Hyttinen, M. Mathioudakis
Evaluating Decision Makers over Selectively Labeled Data: A Causal Modelling Approach
23rd International Conference on Discovery Science, DS 2020.
Best Paper Award.
(Article) (Code)

K. Rantanen, A. Hyttinen, M. Järvisalo
Learning Optimal Cyclic Causal Graphs from Interventional Data
The 10th International Conference on Probabilistic Graphical Models, PGM 2020.
(Article) (Code)

K. Rantanen, A. Hyttinen, M. Järvisalo
Learning Chordal Markov Networks via Stochastic Local Search
Proceedings of the 24th European Conference on Artificial Intelligence, ECAI 2020.
(Article) (Code)

J. Pensar, T. Talvitie, A. Hyttinen, M. Koivisto
A Bayesian Approach for Estimating Causal Effects from Observational Data
Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020.
(Article) (Code)

K. Rantanen, A. Hyttinen, M. Järvisalo
Discovering Causal Graphs with Cycles and Latent Confounders: An Exact Branch-and-Bound Approach
International Journal of Approximate Reasoning, Volume 117, Feb. 2020.
(Article) (Code)

2019

S. Tikka, A. Hyttinen, J. Karvanen
Identifying Causal Effects via Context-specific Independence Relations
Advances in Neural Information Processing Systems 32, NeurIPS 2019.
(Article) (Code)

J. Corander, A. Hyttinen, J. Kontinen, J. Pensar, J. Väänänen
A Logical Approach to Context-Specific Independence
Annals of Pure and Applied Logic, 2019.
(Article) (Implementation)

J. Berg, A. Hyttinen, and M. Järvisalo
Applications of MaxSAT in Data Analysis
EPiC Series in Computing, volume 59, Proceedings of Pragmatics of SAT 2015 and 2018, p. 50- 64, Easychair, 2019.
(Article) (Benchmarks)

2018

K. Rantanen, A. Hyttinen, M. Järvisalo
Learning Optimal Causal Graphs with Exact Search
The 9th International Conference on Probabilistic Graphical Models, PGM 2018.
Best Student Paper Award.
(Article) (Code)

A. Hyttinen, J. Pensar, J. Kontinen, J. Corander
Structure Learning for Bayesian Networks over Labeled DAGs
The 9th International Conference on Probabilistic Graphical Models, PGM 2018.
(Article) (Presentation)

F. Bacchus, A. Hyttinen, M. Järvisalo, and P. Saikko
Reduced Cost Fixing for Maximum Satisfiability
International Joint Conference on Artificial Intelligence, Best Sister Conferences, IJCAI 2018.
(Article) (Implemented e.g. in the MaxHS solver)

2017

K. Rantanen, A. Hyttinen, M. Järvisalo
Learning Chordal Markov Networks via Branch and Bound
Advances in Neural Information Processing Systems 30, NIPS 2017.
(Article) (Code)

A. Hyttinen, S. Plis, M. Järvisalo, F. Eberhardt, and D. Danks
A Constraint Optimization Approach to Causal Discovery from Subsampled Time Series Data
International Journal of Approximate Reasoning, 2017.
(Article) (Presentation) (Code)

F. Bacchus, A. Hyttinen, M. Järvisalo, and P. Saikko
Reduced Cost Fixing in MaxSAT
The 23rd International Conference on Principles and Practice of Constraint Programming, CP 2017.
Distinguished Paper Award.
(Article) (Implemented e.g. in the MaxHS solver)

A. Hyttinen, P. Saikko, and M. Järvisalo
A Core-Guided Approach to Learning Optimal Causal Graphs
International Joint Conference on Artificial Intelligence, IJCAI 2017.
(Article) (Presentation) (Appendix) (Benchmarks)

2016 and earlier

A. Hyttinen, S. Plis, M. Järvisalo, F. Eberhardt, and D. Danks
Causal Discovery from Subsampled Time Series Data by Constraint Optimization
Proceedings of the Eight International Conference on Probabilistic Graphical Models, PGM 2016.
(Article) (Presentation) (Code)

J. Corander, A. Hyttinen, J. Kontinen, J. Pensar, and J. Väänänen
A logical approach to context-specific independence
WoLLIC 2016, 23rd Workshop on Logic, Language, Information and Computation, 2016.
(Article)

A. Hyttinen, F. Eberhardt, and M. Järvisalo
Do-calculus when the True Graph is Unknown
Proceedings of the 31th Conference on Uncertainty in Artificial Intelligence, UAI 2015.
(Article) (Code)

D. Sonntag, M. Järvisalo, Jose M. Pena, A. Hyttinen
Learning Optimal Chain Graphs with Answer Set Programming
Proceedings of the 31th Conference on Uncertainty in Artificial Intelligence, UAI 2015.
(Article) (Code included in the article.)

A. Hyttinen, F. Eberhardt, and M. Järvisalo
Constraint-based Causal Discovery: Conflict Resolution with Answer Set Programming
Proceedings of the 30th Conference on Uncertainty in Artificial Intelligence, UAI 2014.
(Article) (Supplement) (Code)

A. Hyttinen, P. O. Hoyer, F. Eberhardt, and M. Järvisalo
Discovering Cyclic Causal Models with Latent Variables: A General SAT-Based Procedure
Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence, UAI 2013.
(Article) (Presentation) (Code)

A. Hyttinen
Discovering Causal Relations in the Presence of Latent Confounders
Ph.D. Thesis, May, 2013.
The thesis consists of an intro part and the previous six research articles, printed in their original form.
(Intro)

A. Hyttinen, F. Eberhardt, and P. O. Hoyer
Experiment Selection for Causal Discovery
Journal of Machine Learning Research 14(Oct):3041−3071, 2013.
(Article) (Code)

A. Hyttinen, F. Eberhardt, and P. O. Hoyer
Causal Discovery of Linear Cyclic Models from Multiple Experimental Data Sets with Overlapping Variables
Proceedings of the 28th Conference on Uncertainty in Artificial Intelligence, UAI 2012.
(Article) (Appendix) (Presentation) (Code)

A. Hyttinen, F. Eberhardt, and P. O. Hoyer
Learning Linear Cyclic Causal Models with Latent Variables
Journal of Machine Learning Research, 13(Nov):3387-3439, 2012.
(Article) (Code)

A. Hyttinen, F. Eberhardt, and P. O. Hoyer
Noisy-OR Models with Latent Confounding
Proceedings of the 27th Conference on Uncertainty in Artificial Intelligence, UAI 2011.
(Article) (Poster) (Code)

A.Hyttinen, F. Eberhardt, and P. O. Hoyer
Causal discovery for linear cyclic models with latent variables
Proceedings of the Fifth European Workshop on Probabilistic Graphical Models, Helsinki, PGM 2010.
(Article) (Appendix) (Presentation) (Poster) (Code)

P. O. Hoyer and A. Hyttinen
Bayesian Discovery of Linear Acyclic Causal Models
Proceedings of the 25th Conference on Uncertainty in Artificial Intelligence, UAI 2009.
(Article) (Poster) (Code)