SCML2024

Welcome

---

The list of accepted papers is here.

The detailed program is here.

The hybrid conference system is here (the same as the registration system)!

The banquet will start at 6pm (and will end at 8pm) on March 21st at Atrium in Bldg. 1.

In recent years, machine learning methods for scientific computing have attracted much attention. Many methods are a combination of machine learning and/or theories of physics and/or computational mathematics.

This conference aims to showcase the latest research in these areas, which have been fragmented while pursuing research in the same direction, to bridge the gap between them, and to promote collaboration.

Topics will include, but not limited to

• ML for scientific computing
• ML for model discovery
• Physics-informed neural networks
• Operator learning
• Geometric deep learning
• Numerical method for scientific computing
• Discrete mechanics
• Mathematics for ML for science
• Computational algebra for modeling and simulation

This year, this conference will be held in a hybrid format and online participation is possible. Please contact the organizers at yaguchi (at) pearl.kobe-u.ac.jp or scml24-committee (at) geom.jp with any questions.

Conference venue

Kyoto Research Park

134 Chudojiminamimachi, Shimogyoku, Kyoto-shi, Kyoto, 600-8813, Japan

• The conference will take place on the 4th floor of Building 1.
• The reception desk is also located on the the 4th floor of Building 1. Please show the QR code at the reception desk. The QR code is attached to the email sent at the time of registration. You can also download the QR code from the registration system.
• The reception desk will open at 12:00 on 19th March.

Accommodation

We are keeping several rooms in S-Peria Hotel Kyoto (approx. 95,000JPY, from the 18th to the 23rd, breakfast not included) for the participants. This hotel is within walking distance of the conference venue. If you would like to book a room at this hotel, please contact the organizers at yaguchi (at) pearl.kobe-u.ac.jp or scml24-committee (at) geom.jp by March 12th.

Call for papers and submission guidelines

We welcome paper submissions from all related areas with the above topics. This conference will be organized in a similar format to workshops of the major AI conferences, that is,

• Submitted papers will be reviewed by the Program Committee, and all accepted papers will be made available on the conference website. However, authors retain the right to publish elsewhere.
• Submission of papers that are under review or have been recently published in a conference or a journal is allowed.
• Each accepted presentation will be assigned to either an oral presentation or a poster presentation, according to the review reports.
• Submissions should not exceed four pages, excluding references and supplementary materials.
• All submissions must be in the pdf format based on the SCML style file. Please see template.pdf in the SCML style file for other details.

How to submit a paper

Please submit your paper via Microsoft CMT: https://cmt3.research.microsoft.com/SCML2024

Important Dates

• Paper submissions due: January 7, 2024 January 19, 2024 (AoE)
• Notification to authors: January 21, 2024 January 29, 2024 (AoE)

Tutorial speakers

A few tutorials for scientific computing and/or machine learning will be scheduled. The current confirmed speakers are

• Lu Lu (Yale University)

Accurate, efficient, and reliable learning of deep neural operators for multiphysics and multiscale problems



• Takashi Matsubara (Osaka University)

Deep Geometric Mechanics: From Hamiltonian Neural Networks to Discrete-Time Physics and Beyond

Keynote/invited speakers

Several keynote/invited talks will be scheduled. In particular, we are planning invited talks by early and mid-career researchers, as this research field is still at the beginning stage and promoting young researchers is very important. Following is the current tentative list of speakers.

• Christopher J. Budd (Bath University)

Adaptivity and Expressivity in Neural Network Approximations and PINNs

• Elena Celledoni (NTNU)

Deep learning of diffeomorphisms for optimal parametrization of shapes

• Masaaki Imaizumi (University of Tokyo/RIKEN)

Statistics for Modern Data Science: Statistical Analysis on Overparameterized Models and In-Context Learning

• Yuji Nakatsukasa (Oxford University)

Randomized methods for matrix and tensor computations

• Brynjulf Owren (NTNU)

Stability of numerical methods on Riemannian manifolds and applications to neural networks

• Christopher Rackauckas (Massachusetts Institute of Technology)

SciML: Adding Scientific Models as Structure to Improve Machine Learning

• Molei Tao (Georgia Institute of Technology)

Optimization, Sampling, and Generative Modeling in Non-Euclidean Spaces

• Nicolas Boullé (University of Cambridge)

Elliptic PDE learning is provably data-efficient 

• Lena Podina (University of Waterloo)

Universal Physics-Informed Neural Networks and their Applications

• Ayano Kaneda (Waseda University)

Deep Learning Approach to Approximate the Solution for Poisson Matrix

• Shun Sato (The University of Tokyo)

Convergence rates of optimization methods in continuous and discrete time

• Nathanael Bosch (University of Tübingen)

Flexible and Efficient Probabilistic Numerical Solvers for Ordinary Differential Equations

Accepted Papers

The accepted papers include the following papers.

Oral Presentations

• Learning Reduced Order Dynamics via Geometric Representations [paper]

  Imran Nasim (IBM), Melanie Weber (Harvard University)

• Real-time Inference and Extrapolation via a Diffusion-inspired Temporal Transformer Operator (DiTTO) [paper]

  Oded Ovadia (Tel Aviv University), Vivek Oommen (Brown University), Adar Kahana (Brown University), Ahmad Peyvan (Brown University), Eli Turkel (Tel Aviv University), George Em Karniadakis (Brown University)

• Identifying Dynamic Regulation with Adversarial Surrogates [paper]

  Ron Teichner (Technion, Israel Institute of Technology), Naama Brenner (Technion, Israel Institute of Technology) and Ron Meir (Technion, Israel Institute of Technology)

• A Variable Projection Method for Computational PDEs with Artificial Neural Networks [paper]

  Suchuan Dong (Purdue University)

• Towards accurate modeling of dynamics for molecular crystals by scalable variational Gaussian processes [paper]

  Mikhail Tsitsvero (Hokkaido University), Andrey Lyalin (National Institute for Materials Science, Hokkaido University), Mingoo Jin (Hokkaido University)

• Hybrid Modeling Approach Using Cloud Dynamics and Deep Learning for Short-term Solar Forecasting [paper]

  Jun Sasaki (Japan Weather Association), Kenji Utsunomiya (Japan Weather Association), Maki Okada (Japan Weather Association), Koji Yamaguchi (Japan Weather Association)

• Neural Networks are Integrable [paper]

  Yucong Liu (Georgia Institute of Technology)

• Learning phase-space flows using time-discrete implicit Runge-Kutta PINNs [paper]

  Alvaro Fernandez (DESY, Universität Hamburg), Nicolás Mendoza (DESY), Armin Iske (Universität Hamburg), Andrey Yachmenev (DESY, Universität Hamburg), Jochen Küpper (DESY, Universität Hamburg)

Poster Presentations

• Discovering Intrinsic Multi-Compartment Pharmacometric Models Using Physics Informed Neural Networks [paper]

  Imran Nasim (IBM, University of Surrey), Adam Nasim (Merck Group, University of Surrey)

• Using Neural Implicit Flow To Represent Latent Dynamics Of Canonical Systems [paper]

  Imran Nasim (IBM), Joaõ Lucas de Sousa Almeida (IBM)

• Comparing Spectral Bias and Robustness For Two-Layer Neural Networks: SGD vs Adaptive Random Fourier Features [paper]

  Aku Kammonen (KAUST), Lisi Liang (RWTH Aachen), Anamika Pandey (RWTH Aachen), Raúl Tempone (KAUST, RWTH Aachen)

• Synthetic Asset Price Paths Generation Using Denoising Diffusion Probabilistic Model [paper]

  Shujie Liu (University of Waterloo), Justin W.L. Wan (University of Waterloo)

• Efficient Groundwater Flow Modeling Using Deep Neural Operators [paper]

  Maria Luisa Taccari (University of Leeds), He Wang (University College London), Somdatta Goswami (Johns Hopkins University), Mario De Florio (Brown University), Jonathan Nuttall (Deltares), Xiaohui Chen (University of Leeds), Peter K. Jimack (University of Leeds)

• A physics-informed neural network for coupled calcium dynamics in a cable neuron [paper]

  Zachary M. Miksis (Temple University), Gillian Queisser (Temple University)

• Learning Hamiltonian dynamics Under Uncertainty via Symplectic Gaussian Processes [paper]

  Yusuke Tanaka (NTT)

• SPIGAN: A Generative Adversarial Network Supervised by Sparse Identification to Learn Governing Equations from Scarce Data [paper]

  Yue Shen (The University of Tokyo), Chen Yu (The University of Tokyo)

• Domain-Decomposed Physics-Informed Neural Network Prediction on Cartesian CFD Framework [paper]

  Takashi Misaka (National Institute of Advanced Industrial Science and Technology (AIST)), Yusuke Mizuno (National Institute of Advanced Industrial Science and Technology (AIST)), Shogo Nakasumi (National Institute of Advanced Industrial Science and TechnoTechnology (AIST)), Yoshiyuki Furukawa (National Institute of Advanced Industrial Science and Technology (AIST))

• Sparse Representation of Koopman Operator [paper]

  Yuya Note (Kobe University), Takaharu Yaguchi (Kobe University), Toshiaki Omori (Kobe University)

• Biologically plausible local synaptic learning rules implement CNNs and denoising autoencoders [paper]

  Masataka Konishi (Kwansei Gakuin University), Keiji Miura (Kwansei Gakuin University)

• Toward Bayesian Deep Grey-box Modeling [paper]

  Naoya Takeishi (The University of Tokyo, RIKEN)

• Synthetic label masks mapped on ocean satellite background for oil seepage detection [paper]
  Lionel Boillot (TotalEnergies), Frédérik Pivot (TotalEnergies), Félix Klein (TotalEnergies)

• Physics Informed Neural Networks with Application in Computational Structural Mechanics [paper]

  Aryan Verma (The Ohio State University), Dineshkumar Harursampath (Indian Institute of Science), Rajnish Mallick (Thapar Institute of Engineering & Technology), Prasant Sahay (Indian Institute of Science), Krishna Kant Mishra (Indian Institute of Science)

Registration

Registration is required for you to participate in the conference. In particular, to present your work, at least one of the authors should make a registration.

The system for registration will be available at https://scml2024.award-con.com

Registration fee

• Early registration (until February 19, 2024)
  • Regular participant: 50000JPY
  • Regular participant (online participation only): 40000JPY
  • Student participant: 30000JPY
  • Student participant (online participation only): 20000JPY
    
• Regular registration
  • Regular participant: 60000JPY
  • Regular participant (online participation only): 40000JPY
  • Student participant: 40000JPY
  • Student participant (online participation only): 20000JPY

Organizers

• Takaharu Yaguchi (Kobe University)
• Hiroaki Yoshimura (Waseda University)
• Nobuki Takayama (Kobe University)
• Toshiaki Omori (Kobe University)
• Takashi Matsubara (Osaka University)
• Kumiko Hori (National Institute for Fusion Science)
• Saiei-Jaeyeong Matsubara-Heo (Kumamoto University)
• Mizuka Komatsu (Kobe University)
• Yuhan Chen (Kobe University)
• Baige Xu (Kobe University)