Als u wijzigingen in Pure hebt gemaakt, zullen deze hier binnenkort zichtbaar zijn.

Persoonlijk profiel

Quote

“Conjecturally, in the depths of the human brain runs an immensely powerful, simple, efficient and task- and signal-independent learning algorithm. It is my ultimate aim to use mathematics to uncover and develop such algorithms.”

Research profile

Jim Portegies is an Assistant Professor in the Applied Analysis group of the Centre for Analysis, Scientific computing and Applications (CASA) at Eindhoven University of Technology (TU/e). Jim Portegies works in the mathematical fields of analysis, measure theory and geometry, and applies techniques from these fields to problems in machine-learning and artificial intelligence. He has applied techniques from spectral geometry to prove guarantees on performance of nonlinear dimensionality reduction algorithms. Currently, he is investigating how to design algorithms that mimic how humans and animals learn.  

Despite the large number of recent advances in artificial intelligence, humans still outperform machines in many tasks. The central question of how to design machines that learn like humans is still wide open. The answer may lie in universal learning algorithms. Such algorithms are simple, efficient and can be applied to a broad variety of signals and tasks and are conjectured to exist in the depths of the human brain.

Academic background

Jim Portegies obtained his MSc in Industrial and Applied Mathematics and Applied Physics from the TU/e in 2009. He spent the 2007-2008 academic year as an exchange student at the University of Bonn, Germany. He received his PhD in Mathematics from the Courant Institute of Mathematical Sciences in New York. In the Fall of 2013, he spent a semester at NYU Shanghai, in Shanghai, China. After completing his PhD in 2014, he spent two years a postdoc at the Max Planck Institute for Mathematics in the Sciences in Leipzig, Germany until he returned to the TU/e as an assistant professor in Mathematics in 2016. Jim is a member of the TU/e Young Academy of Engineering. 

Vingerafdruk Duik in de onderzoeksthema's waar Jim W. Portegies actief is. Deze onderwerplabels komen voort uit het werk van deze persoon. Samen vormen ze een unieke vingerafdruk.

trimers Fysica en Astronomie
harmonics Fysica en Astronomie
Gromov-Hausdorff Convergence Rekenkunde
Equipartition Rekenkunde
Nonlinear Eigenvalue Rekenkunde
Eigenvalue Rekenkunde
Ricci Curvature Rekenkunde
Riemannian Manifold Rekenkunde

Netwerk Recente externe samenwerking op landenniveau. Duik in de details door op de stippen te klikken.

Onderzoeksoutput 2008 2019

  • 37 Citaten
  • 12 Tijdschriftartikel
  • 2 Rapport
  • 2 Conferentiebijdrage
  • 1 Paper

Can VAEs capture topological properties?

Perez Rey, L. A., Menkovski, V. & Portegies, J. W., 2019.

Onderzoeksoutput: Bijdrage aan congresPaper

Brownian movement
Topology
3 Downloads (Pure)

Total variation and mean curvature PDEs on the space of positions and orientations

Duits, R., St-Onge, E., Portegies, J. & Smets, B., 5 jun 2019, Scale Space and Variational Methods in Computer Vision - 7th International Conference, SSVM 2019, Proceedings. Lellmann, J., Modersitzki, J. & Burger, M. (redactie). Berlin: Springer, blz. 211-223 13 blz. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); vol. 11603 LNCS).

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

Total Variation
Mean Curvature
Magnetic resonance imaging
Mean Curvature Flow
Lie groups
27 Downloads (Pure)

Asymptotic dependency structure of multiple signals: Asymptotic equipartition property for diagrams of probability spaces

Matveev, R. & Portegies, J. W., dec 2018, In : Information Geometry. 1, 2, blz. 237-285

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

Open Access
Bestand
Equipartition
Probability Space
Diagram
Asymptotically equivalent
Neuroscience
3 Downloads (Pure)

Continuity of nonlinear eigenvalues in CD (K, ∞) spaces with respect to measured Gromov–Hausdorff convergence

Ambrosio, L., Honda, S. & Portegies, J. W., 1 apr 2018, In : Calculus of Variations and Partial Differential Equations. 57, 2, 34.

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

Nonlinear Eigenvalue
K-space
Eigenvalue
Metric Measure Space
Laplace Operator

Ergo learning

Portegies, J. W., sep 2018, In : Nieuw Archief voor Wiskunde. 5th Series, Volume 19, 3, blz. 199-205

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademic

Cursussen

Analysis 1

1/09/12 → …

Cursus

Analysis 2

1/09/12 → …

Cursus

Data analytics for engineers

1/09/17 → …

Cursus

Linear algebra and applications

1/09/17 → …

Cursus

Scriptie

A comparison of the Krasnoselskii spectrum and the homotopy significant spectrum

Auteur: Fokma, S. (., 6 jul 2018

Begeleider: Portegies, J. (Afstudeerdocent 1)

Scriptie/masterproef: Bachelor

Bestand

Active learning in VAE latent space

Auteur: Tonnaer, L., 25 sep 2017

Begeleider: Menkovski, V. (Afstudeerdocent 1), Portegies, J. W. (Afstudeerdocent 2) & Holenderski, M. (Afstudeerdocent 2)

Scriptie/masterproef: Master

Bestand

Computer programs for analysis

Auteur: Beurskens, T. P., 1 jul 2019

Begeleider: Portegies, J. W. (Afstudeerdocent 1)

Scriptie/masterproef: Bachelor

Bestand

Convergence of several reinforcement learning algorithms

Auteur: Mohamed, A., 29 okt 2018

Begeleider: Portegies, J. W. (Afstudeerdocent 1)

Scriptie/masterproef: Bachelor

Bestand

Differential equations driven by rough signals

Auteur: Verstraelen, L. H., 31 aug 2018

Begeleider: Prokert, G. (Afstudeerdocent 1) & Portegies, J. (Afstudeerdocent 2)

Scriptie/masterproef: Bachelor

Bestand