Description
Leader of the group Machine Learning: Peter Grünwald.
Our research group focuses on how computer programs can learn from and understand data, and then make useful predictions based on it. These algorithms integrate insights from various fields, including statistics, artificial intelligence and neuroscience.
Machine-learning applications are increasingly part of every aspect of life, from speech recognition on cell phones to illness prediction in healthcare. One common problem is extremely polluted data, for which no single model can provide adequate explanations. At CWI we address this issue with statistical machine learning based on combining predictions from different models and experts in order to achieve reliable conclusions.
We also study how networks of neurons in the brain process information, and how modern deep-learning methods can benefit from neuroscience. We develop novel neural networks, like Deep Adaptive Spiking Neural Networks, and also theoretical models of neural learning and information processing in biology. Applications of our work range from low-energy consumption neural machine learning to neuroprosthetics, to increased insight into the question of how the brain works.
News
Researchers develop neural model for working memory
Neuroscientists of Centrum Wiskunde & Informatica (CWI) and the Netherlands Institute for Neuroscience (NIN) have developed a biologically plausible neural network model that can learn to remember past events in order to use them in the future. The researchers developed their model by combining theoretical principles from machine learning with insights from neuroscience.
CWI starts research on spiking neural networks
Sander Bohte, researcher in the life science group of the Centrum Wiskunde & Informatica (CWI) in Amsterdam, starts in collaboration with researchers from the University of Amsterdam (UvA) a new project on spiking neural networks.
Brain mechanisms better understood with new model
Building a neural network with the same properties and capacity as the human brain is the holy grail in neuroinformatics. Such a network would not only explain the inner workings of the brain, but would also pave the road for brain-controlled machines such as computers operated by thought and robot limbs for the handicapped.
CWI simulates brain activity on video cards
Neuroinformaticists of Centrum Wiskunde & Informatica (CWI) in Amsterdam managed to simulate complex brain activity on simple video cards. The simulated brain contains 50,000 neurons communicating with 35 million signals per second. This is comparable to the brain capacity of insects such as ants or flies.
Current events
Event-based Asynchronous Neuro-Cognitive Control
- 2019-08-26T12:00:00+02:00
- 2019-08-28T15:00:00+02:00
Event-based Asynchronous Neuro-Cognitive Control
Start: 2019-08-26 12:00:00+02:00 End: 2019-08-28 15:00:00+02:00
Centrum Wiskunde & Informatica, Amsterdam (26 – 28 August 2019)
Compared to modern Deep Learning, the human brain needs vastly fewer examples to learn tasks and is massively more energy efficient. The brain’s ability to control, and learn to control, hundreds of flexible and variable muscles for motion remains unsurpassed. Sensory information processing, cognitive deliberation and subsequent muscle control is also an ongoing continuous-time process, a fact mostly ignored in deep learning models. Finally, the brain solves an inherently multi-scale problem: fast high-dimensional sensory information feeds, (relatively) slow and low dimensional cognitive deliberations, which are then translated into fast and high-dimensional muscle activations resulting in typically low-dimensional feedback for learning from the environment in the form of success or failure of actions. Recently, models for a number of components in this sensori-cognitive-motor chain have been proposed.
In this workshop, we aim to bring together experts in computational neuroscience and machine learning to foster collaboration and start working on integrating the various components into coherent end-to-end cognitive models. Specifically, we aim to determine the `missing components’ that are needed to implement fast and effective sensori-cognitive-motor models.
At a high level, the main topics in the workshop include End-to-end time-continuous learning, Multi-level Motion Representation, Planning and Control and Event-based asynchronous neural computation.
The workshop will be held over a 3-day period at CWI, the Dutch National Centre for Mathematics and Computer Science. Invited speakers will bring broad-ranging expertise from neuroscience and machine learning, with a special focus on efficient platforms. The workshop will also have plenty of time for formal and informal discussions.
Confirmed invited speakers
Thomas Nowotny (Univ Sussex, UK)
Wulfram Gerstner (EPFL, CH)
Shih-Chii Liu (INI Zurich, CH)
Eleni Vasilaki (Univ Sheffield, UK)
Friedemann Zenke (FMI Basel, CH)
Terry Stewart (Univ Waterloo, CA)
De Ma (Zhejiang University,CN)
Chiara Bartolozzi (IIT Genova, IT)
Raoul-Martin Memmesheimer (Univ Bonn, DE)
Mihai Petrovici (Univ Heidelberg, DE)
Jamie Knight (Univ Sussex, UK)
Participation and deadlines
Participation in the workshop is free of charge and includes the lunches for the three days. However, the workshop is limited to a restricted number of people given the capacity of the venue and the aim of the event. Prospective attendees should express their interest in participating by submitting the form downloadable from the link below, and their willingness to contribute a poster. For the poster, people are also invited to submit a short/extended abstract (max 2 pages), possibly including figures. The submissions will be selected on the basis of the quality and of how well they fit the workshop’s aims.
Both senior and young researchers, including Ph.D. students, are encouraged to submit to ensure a full representation of the community.
Deadline: 11.07.2019
Please use this form and submit it and the optional abstract (subject: “EANCC submission: name, last name”) to: eancc2019@gmail.com
Organising Committee
- Sander Bohte – CWI / Univ. of Amsterdam / RUG Groningen, NL
- Aditya Gilra – IST Austria, AT
- Qinghai Guo – Huawei Research, CN
- J. Camilo Vasquez Tieck – FZI Karlsruhe, GE
- Email: Sander Bohte
- Phone: +31-205924074
-
iCal
Members
Associated Members
Publications
-
Yin, B, Balvert, M, van der Spek, R.A.A, Dutilh, B.E, Bohte, S.M, Veldink, J, & Schönhuth, A. (2019). Using the structure of genome data in the design of deep neural networks for predicting amyotrophic lateral sclerosis from genotype. In Bioinformatics (Vol. 35, pp. i538–i547). doi:10.1093/bioinformatics/btz369
-
Een computer kan niet alles leren - NRC- 1 juni 2019. (2019). Een computer kan niet alles leren - NRC- 1 juni 2019.
-
van Doorn, J, Ly, A, Marsman, M, & Wagenmakers, E.-J. (2019). Bayesian estimation of Kendall's τ using a latent normal approach. Statistics & Probability Letters, 145, 268–272. doi:10.1016/j.spl.2018.10.004
-
Zambrano, D, Nusselder, R.B.P, Scholte, H.S, & Bohte, S.M. (2019). Sparse computation in adaptive spiking neural networks. Frontiers in Neuroscience, 12(JAN). doi:10.3389/fnins.2018.00987
-
Kaufmann, E, Koolen-Wijkstra, W.M, & Garivier, A. (2018). Sequential test for the lowest mean: From Thompson to Murphy sampling. In Advances in Neural Information Processing Systems (pp. 6332–6342).
-
Karamanis, M, Zambrano, D, & Bohte, S.M. (2018). Continuous-time spike-based reinforcement learning for working memory tasks. In Lecture Notes in Computer Science/Lecture Notes in Artificial Intelligence (pp. 250–262). doi:10.1007/978-3-030-01421-6_25
-
Dora, S, Pennartz, C, & Bohte, S.M. (2018). A deep predictive coding network for inferring hierarchical causes underlying sensory inputs. In Lecture Notes in Computer Science/Lecture Notes in Artificial Intelligence (pp. 457–467). doi:10.1007/978-3-030-01424-7_45
-
Pozzi, I, Nusselder, R.B.P, Zambrano, D, Bohte, S.M, & Iliadis, L. (2018). Gating sensory noise in a spiking subtractive LSTM. In V Kůrková, Y Manolopoulos, B Hammer, & I Maglogiannis (Eds.), Proceedings of Artificial Neural Networks and Machine Learning - ICANN 2018 (pp. 284–293). Springer, Cham. doi:10.1007/978-3-030-01418-6_28
-
Grünwald, P.D, & de Heide, R. (2018). Invited discussion to the paper Using Stacking to Average Bayesian Predictive Distributions by Yao, Vehtari, Simpson and Gelman. Bayesian Analysis, 13(3), 917–1003.
-
Yin, B, Balvert, M, Zambrano, D, Schönhuth, A, & Bohte, S.M. (2018). An image representation based convolutional network for DNA classification. In 6th International Conference on Learning Representations.
Software
Squint: Experimenting in Prediction with Expert Advice problems
Squint provides a codebase to perform numerical proof-of-concept experiments in learning theory, particularly in Prediction with Expert Advice problems, a core problem in learning theory.
Current projects with external funding
-
Efficient Deep Learning Platforms (eDLP)
-
Enabling Personalized Interventions (EPI)
-
Safe Bayesian Inference: A Theory of Misspecification based on Statistical Learning (SAFEBAYES)
-
Spiking Neural Networks research program
Related partners
-
Philips
-
KPMG
-
SURFsara B.V.
-
Technische Universiteit Eindhoven
-
Universiteit Twente
-
Universiteit van Amsterdam
-
Vrije Universiteit Amsterdam