Bridging scales: At the crossroads among renormalisation group, multi-scale modelling, and deep learning

15 Apr 2024, 08:30 → 19 Apr 2024, 19:00 Europe/Rome

Aula Renzo Leonardi (ECT*)

Alessandro Roggero, Francesco Pederiva (University of Trento and INFN-TIFPA), Raffaello Potestio (UniTn - Physics Dept.), Roberto Menichetti (Physics Department, University of Trento)

***WARNING: for hotel bookings do not give your data to any external services upon email request; there are known cases of fraud***

Machine learning will define the 21st Century: from simple image classification to text generation and decision making, its impact on society will be nothing but immense. At present, the detailed mechanisms behind the power of AI still evade our understanding; growing evidence, however, suggests that it is possible to rationalise how deep learning works in terms that are very familiar to theoretical physicists, that is, the renormalisation group. The systematic, hierarchical coarsening of detailed information into increasingly simpler and more collective features is a cornerstone of modern physics, and it can be leveraged not only to make sense of machine learning’s baffling
capabilities, but also and most importantly to steer its development. This workshop will explore the area where theoretical physics of soft and condensed matter and deep learning overlap, looking for novel and more powerful tools to model, investigate, and understand the world around us.

Monday 15 April

1 Registration

2 Welcome

3 Topological polymeric soft materials: Some examples

Speaker: Luca Tubiana (University of Trento, Italy)

10:40 Coffee break

4 Renormalisation group techniques for polymers on fractal lattices

Speaker: Antonio Trovato (University of Padova - Department of Physic and Astronomy)

trovato_ect2024.pdf

12:20 Lunch

5 Coupling DFT, SSCHA and SchNet NN for Hydrogen desorption temperature calculations in Magnesium Hydride Nanoclusters

Speaker: Andrea Pedrielli (FBK)

Presentazione.pdf

15:30 Coffee break

Open discussion

Tuesday 16 April

6 Fundamental limits of Shallow Neural Networks for Supervised Learning

Speaker: Jean Barbier (ICTP)

FundamentalLimitsNN (1).pdf

7 Modelling complex data with RBMs

Speaker: Beatriz Seoane Bartolome (LISN, Paris-Saclay)

ECTrento_BSeoane.pdf

10:40 Coffee break

8 Renormalizing Molecular Dynamics

Speaker: Pietro Faccioli (University of Milan-Bicocca)

12:20 Lunch

9 Natural swarms in 3.99 dimensions

Speaker: Mattia Scandolo (ISC-CNR)

scandolo_ect.pdf

10 A quantum machine learning perspective on phase transitions

Speaker: Oriel Kiss (CERN)

Briding_scales24_Kiss.pptx

15:30 Coffee break

11 Operator Leaning Renormalization Group for Quantum Simulations

Speaker: Di Luo (Massachusetts Institute of Technology)

Open discussion

Social Dinner

Wednesday 17 April

12 Machine learning in biomolecular simulations: from characterizing conformational free energy landscapes to scale bridging

Speaker: Christine Peter (University of Konstanz)

Trento_2024_CPeter.pdf

13 What do Transformers learn when trained via Masked Language Modelling?

Speaker: Federica Gerace

Transformers_Potts_Federica-Gerace_ECT-Trento.pdf

10:40 Coffee break

14 Inferring phase transitions and critical exponents from limited observations with Thermodynamic Maps

Speaker: Pratyush Tiwary (University of Maryland, College Park)

tiwary_trento_apr2024.pdf

tiwary_trento_apr2024.pptx

12:20 Lunch

15 Loss is more: exploring the weight space of a perceptron via enhanced sampling techniques

Speaker: Margherita Mele (University of Trento)

15:30 Coffee break

16 Diffusion Model as Stochastic Quantization in Lattice QFT

Speaker: Kai Zhou (Frankfurt Institute for Advanced Studies)

Trento_DMSQ_Kai.pdf

Open discussion

Thursday 18 April

17 A computation-dissipation tradeoff for machine learning at the mesoscale

Speaker: Alessandro Ingrosso (The Abdus Salam International Centre for Theoretical Physics)

ect_trento.pdf

18 Exploring the landscape of model representations

Speaker: Will Noid (penn state)

ECTstar-ModelLandscape-18April2024.pptx

10:40 Coffee break

19 Neural Quantum States for Quantum Many-Body Physics

Speaker: Arnau Rios Huguet (Institute of Cosmos Sciences, University of Barcelona, Spain)

rios_ML_Trento_24.pdf

12:20 Lunch

20 Thrustworthy machine learning for materials

Speaker: Federico Grasselli (EPFL)

Trustworthy_ML_Trento.pdf

21 Tracking and Analysis of Active Droplet Dynamics

Speaker: Matteo Scandola (University of Trento)

ETC_Matteo Scandola.pdf

15:30 Coffee break

22 Transferable coarse-grained models accelerate chemical-space exploration

Speaker: Tristan Bereau (Heidelberg University)

202404_trento_bereau.pdf

23 TBA

Speaker: Yannick Meurice (University of Iowa)

Open discussion

Social Dinner

Friday 19 April

24 Fixed point actions from lattice gauge equivariant convolutional neural networks

Speaker: Andreas Ipp (TU Wien)

trento_ipp_2024_v16.pdf

25 Towards multi-scale models of active filaments

Speaker: Emanuele Locatelli (University of Padova)

Locatelli_Trento_2024.pdf

10:40 Coffee break

26 An integrated photonic neuromorphic interface for dimensionality expansion of optical signals

Speaker: Alessio Lugnan (University of Trento)

BrisgingScales_ECT_workshop_19Apr2024.pdf

12:20 Lunch

Departure