Description An ICTP Hybrid Meeting

“Youth in high dimensions” gives to young researchers from academia and industry the opportunity to gather and present recent results on high-dimensional statistical problems, in areas ranging from machine learning and inference to neuroscience or statistical physics.
 
Typical problems in high-dimensions include estimating noisy signals (compressed sensing, PCA and tensor decomposition), analysing (deep) neural-networks, detecting communities in large networks, or modelling disordered spin systems. These seemingly unrelated problems share many similarities in their phenomenology and in the tools used to analyse them. The core of the field is made of a very active, diverse and quickly expanding community of physicists, computer scientists, mathematicians, information theorists and engineers, with the common desire to tackle increasingly challenging problems at the forefront of data science. This conference aims at reinforcing the links among this interdisciplinary community, and in particular of its youngest theory-oriented members, and to bring forward the latest development happening in the high-dimensional world.
 
Topics:
  • High-dimensional statistics
  • Statistical physics
  • Machine learning 
  • Complex systems 
  • Random matrix theory
  • Computational neuroscience

Speakers:
A. ALTIERI, Université Paris Cité, France
A. BIETTI, NYU, USA
E. CORNACCHIA, EPFL, Switzerland
N. FLAMMARION, EPFL, Switzerland
L. FOINI, CNRS IPhT Université Paris-Saclay, France
M. GEIGER, MIT, USA
J. KADMON, The Hebrew University, Israel
J. KO, ENS Lyon, France
M. LINDSEY, NYU, USA
J. LONG, Princeton University, USA
B. MCKENNA, IST, Austria
C. PEHLEVAN, Harvard University, USA
N. RAZIN, Tel Aviv University, Israel
D. SCHRÖDER, ETH, Switzerland
M. SELLKE, Stanford University, USA
G. SICURO, King’s College London, UK
C. THRAMPOULIDIS, University of British Columbia, USA
Y. WEI, University of Pennsylvania, USA
F. YANG, ETH, Switzerland
Y. ZHU, University of California, USA
 

Please note in the application form you will be asked to submit a poster abstract. A number of abstracts will be selected for the poster session.

Grants: A limited number of grants are available to support the attendance of selected participants, with priority given to participants from developing countries. There is no registration fee.

In person participation: As regards the COVID-19 policy, we advise to follow the updated rules available on the ICTP page Access Guidelines for Visitors.
 
Go to day
  • Monday, 27 June 2022
    • 08:30 - 09:00
      Location: Leonardo Building - Lobby
      • 08:30 Registration formalities (in person participants) 25'
        REGISTRATION: Upon arrival, Visitors not staying in the ICTP Guest House, are kindly requested to complete registration formalities at the
        Leonardo Building Lobby from 8.30 till 08:55.
      • 08:55 Opening remarks 5'
    • 09:00 - 12:30 ML Theory 1 & RMT 1
      Location: Leonardo Building - Budinich Lecture Hall
      • 09:00 An Initial Alignment between Neural Network and Target is Needed for Gradient Descent to Learn 30'
        Speaker: Elisabetta CORNACCHIA (EPFL, Switzerland)
        Material: Video
      • 09:30 The Role of Stochasticity in Learning Algorithms 30'
        /Remote/
        Speaker: Nicolas FLAMMARION (EPFL, Switzerland)
        Material: Video
      • 10:00 Minimum L1-norm Interpolators: Precise Asymptotics and Multiple Descent 30'
        /Remote/
        Speaker: Yuting WEI (University of Pennsylvania, USA)
        Material: Video
      • 10:30 Coffee break 30'
      • 11:00 Landscape Complexity Beyond Invariance 30'
        Speaker: Ben MCKENNA (IST, Austria)
        Material: Video
      • 11:30 Random Matrix Resolvent Analysis via Cumulant Expansion 30'
        Speaker: Dominik SCHROEDER (ETH Zurich, Switzerland)
        Material: Video
    • 14:00 - 17:30
      Location: Leonardo Building - Budinich Lecture Hall
      • 14:00 Poster session in person 2h0'
      • 16:00 Coffee break 30'
      • 16:30 Discussion 1h0'
  • Tuesday, 28 June 2022
    • 09:00 - 12:30 Glasses & Beyond
      Location: Leonardo Building - Budinich Lecture Hall
      • 09:00 Dynamical Mean-Field Theory: from Glassy Systems to Ecology and Inference 30'
        Speaker: Ada ALTIERI (Paris Diderot, France)
        Material: Video
      • 09:30 The Planted Matching Problem and its Variations 30'
        Speaker: Gabriele SICURO (King’s College London, UK)
        Material: Video
      • 10:00 Annealed Averages in Spin and Matrix Models 30'
        Speaker: Laura FOINI (CNRS IPhT Université Paris-Saclay, France)
        Material: Video
      • 10:30 Coffee break 30'
      • 11:00 Discussion 1h30'
    • 14:00 - 16:30 ML 4 Sciences
      Location: Leonardo Building - Budinich Lecture Hall
      • 14:00 Perturbational Complexity by Distribution Mismatch: A Systematic Analysis of Reinforcement Learning in Reproducing Kernel Hilbert Space 30'
        /Remote/
        Speaker: Jihao LONG (Princeton University, USA)
        Material: Video
      • 14:30 Group Theory for Machine Learning 30'
        Speaker: Mario GEIGER (MIT, USA)
        Material: Video
      • 15:00 Thermal State Sampling for Numerical Linear Algebra 30'
        Speaker: Michael LINDSEY (NYU, USA)
        Material: Video
      • 16:00 Hike all together 30' ( Outside )
  • Wednesday, 29 June 2022
    • 09:00 - 12:30 ML Theory 2
      Location: Leonardo Building - Budinich Lecture Hall
      • 09:00 Fast Rates for Noisy Interpolation Require Rethinking the Effects of Inductive Bias 30'
        Speaker: Fan YANG (ETH Zurich, Switzerland)
        Material: Video
      • 09:30 Finding Structures in Large Models: Imbalance Trouble 30'
        Speaker: Christos THRAMPOILIDIS (University of British Columbia, USA)
        Material: Video
      • 10:00 A Universal Law of Robustness via Isoperimetry 30'
        Speaker: Mark SELLKE (Stanford University, USA)
        Material: Video
      • 10:30 Group photo 'virtual' 2'
      • 10:32 Group photo 'in person' 3'
      • 10:35 Coffee break 25'
      • 11:00 Discussion 1h30'
    • 14:00 - 18:00 Theoretical Neuro
      Location: Leonardo Building - Budinich Lecture Hall
      • 14:00 Deep Learning Theory at Limits 30'
        Speaker: Cengiz PEHLEVAN (Harvard University, USA)
        Material: Video
      • 14:30 Order from Chaos: Computation and Learning in Cortical Networks 30'
        Speaker: Jonathan KADMON (The Hebrew University, Israel)
        Material: Video
      • 16:00 Coffee break 30'
      • 16:30 Discussion 1h30'
  • Thursday, 30 June 2022
    • 09:00 - 12:30 RMT 2 & ML Theory 3
      Location: Leonardo Building - Budinich Lecture Hall
      • 09:00 Matrix Estimation Problems 30'
        Speaker: Justin KO (ENS Lyon, France)
        Material: Video
      • 09:30 Non-backtracking Spectral Clustering in Sparse Random Hypergraphs 30'
        Speaker: Yizhe ZHU (University of California, USA)
        Material: Video
      • 10:30 Coffee break 30'
      • 11:00 Generalization in Deep Learning Through the Lens of Implicit Rank Lowering 30'
        /Remote/
        Speaker: Noam RAZIN (Tel Aviv University, Israel)
        Material: Video
      • 11:30 Learning Single-Index Models with Shallow Neural Networks 30'
        Speaker: Alberto BIETTI (NYU, USA)
        Material: Video