Publications associées à Neural networks trained with SGD learn distributions of increasing complexity

Publications associées (21)

Machine learning enhanced atom probe tomography analysis

Atom probe tomography (APT) is a burgeoning characterization technique that provides compositional mapping of materials in three-dimensions at near-atomic scale. Since its significant expansion in the past 30 years, we estimate that one million APT dataset ...

2026

Enhancing Lift and Dynamic Stability Through Morphological Optimization of Wings for Tailless Flapping-Wing Robots

Josephine Anna Eleanor Hughes

The design of wings is crucial for tailless Flapping-wing robots (FWRs), as these robots rely exclusively on their wings for lift generation and body control. However, the current methodology for wing design primarily depends on the experience and heuristi ...

2025

Hadamard product in deep learning: Introduction, Advances and Challenges

Volkan Cevher, Yongtao Wu

While convolution and self-attention mechanisms have dominated architectural design in deep learning, this survey examines a fundamental yet understudied primitive: the Hadamard product. Despite its widespread implementation across various applications, th ...

2025

A High Dimensional Statistical Model for Adversarial Training: Geometry and Trade-Offs

Florent Gérard Krzakala, Bruno Loureiro, Matteo Vilucchio

This work investigates adversarial training in the context of margin-based linear classifiers in the high-dimensional regime where the dimension d and the number of data points n diverge with a fixed ratio α = n/d. We introduce a tractable mathematical mod ...

2025

Robustly Learning Regions of Attraction From Fixed Data

Colin Neil Jones

While stability analysis is a mainstay for control science, especially computing regions of attraction of equilibrium points, until recently most stability analysis tools always required explicit knowledge of the model or a high-fidelity simulator represen ...

2025

Representing and machine learning complex electronic-structure properties of materials

Austin Jacob Zadoks

The rise of high-throughput first-principles calculations of materials since the turn of the millennium has gifted the fields of physics, quantum chemistry, and materials science with a mountainous and ever-growing pile of data ready to be mined for hidden ...

EPFL2025

Monitoring of water volume in a porous reservoir using seismic data: A 3D simulation study

Jan Sickmann Hesthaven

A potential framework to estimate the volume of water stored in a porous storage reservoir from seismic data is neural networks. In this study, the man-made groundwater reservoir is modeled as a coupled poroviscoelastic–viscoelastic medium, and the underly ...

2024

Higher-order knowledge-enhanced recommendation with heterogeneous hypergraph multi-attention

Thanh Trung Huynh

Recent advancements in recommender systems have focused on integrating knowledge graphs (KGs) to leverage their auxiliary information. The core idea of KG-enhanced recommenders is to incorporate rich semantic information for more accurate recommendations. ...

2024

LiDAR-NeRF: Novel LiDAR View Synthesis via Neural Radiance Fields

Mathieu Salzmann

We introduce a new task, novel view synthesis for LiDAR sensors. While traditional model-based LiDAR simulators with style-transfer neural networks can be applied to render novel views, they fall short of producing accurate and realistic LiDAR patterns bec ...

Association for Computing Machinery, Inc2024

Positive-Unlabeled Constraint Learning for Inferring Nonlinear Continuous Constraints Functions From Expert Demonstrations

Baiyu Peng, Aude Billard

Planning for diverse real-world robotic tasks necessitates to know and write all constraints. However, instances exist where these constraints are either unknown or challenging to specify accurately. A possible solution is to infer the unknown constraints ...

2024

Function Learning with Financial Applications

Paolo Colusso

This thesis sets itself the goal of investigating function learning approaches. The topic is extremely relevant today, since the abundance of data, jointly with the technological progress which has enhanced computing power, now allow to represent functions ...

EPFL2024

Rigorous Dynamical Mean-Field Theory for Stochastic Gradient Descent Methods

Lenka Zdeborová, Florent Gérard Krzakala, Emanuele Troiani

We prove closed-form equations for the exact high-dimensional asymptotics of a family of first-order gradient-based methods, learning an estimator (e.g., M-estimator, shallow neural network) from observations on Gaussian data with empirical risk minimizati ...

Society for Industrial & Applied Mathematics (SIAM)2024

Unsupervised quality monitoring of metal additive manufacturing using Bayesian adaptive resonance

Patrik Willi Hoffmann

Metal additive manufacturing is a recent breakthrough technology that promises automated production of complex geometric shapes at low operating costs. However, its potential is not yet fully exploited due to the low reproducibility of quality in mass prod ...

CELL PRESS2024

Enhancing Explainability and Performance in AI-Driven Face Recognition and Deepfake Detection

Yuhang Lu

The rapid advancement of artificial intelligence (AI) has greatly influenced numerous research areas, leading to significant breakthroughs in human face-related technologies, particularly in face recognition and deepfake detection. While offering substanti ...

EPFL2024

Fast dynamic 1D simulation of divertor plasmas with neural PDE surrogates

Yoeri Renier Johannes Poels

Managing divertor plasmas is crucial for operating reactor scale tokamak devices due to heat and particle flux constraints on the divertor target. Simulation is an important tool to understand and control these plasmas, however, for real-time applications ...

IOP Publishing Ltd2023

The automated collapse data constructor technique and the data-driven methodology for seismic collapse risk assessment

Dimitrios Lignos, Alexandre Massoud Alahi, Nenad Bijelic

Majority of the past research on application of machine learning (ML) in earthquake engineering focused on contrasting the predictive performance of different ML algorithms. In contrast, the emphasis of this paper is on the use of data to boost the predict ...

WILEY2023

AMAE: Adaptation of Pre-trained Masked Autoencoder for Dual-Distribution Anomaly Detection in Chest X-Rays

Jean-Philippe Thiran, Seyedbehzad Bozorgtabar

Unsupervised anomaly detection in medical images such as chest radiographs is stepping into the spotlight as it mitigates the scarcity of the labor-intensive and costly expert annotation of anomaly data. However, nearly all existing methods are formulated ...

Springer International Publishing Ag2023

From tensor-network quantum states to tensorial recurrent neural networks

Giuseppe Carleo, Filippo Vicentini, Riccardo Rossi, Dian Wu

We show that any matrix product state (MPS) can be exactly represented by a recurrent neural network (RNN) with a linear memory update. We generalize this RNN architecture to two-dimensional lattices using a multilinear memory update. It supports perfect s ...

AMER PHYSICAL SOC2023

Measuring Complexity of Learning Schemes Using Hessian-Schatten Total Variation

Michaël Unser, Shayan Aziznejad, Joaquim Gonçalves Garcia Barreto Campos

2023

Training Provably Robust Models by Polyhedral Envelope Regularization

Sabine Süsstrunk, Mathieu Salzmann, Chen Liu

Training certifiable neural networks enables us to obtain models with robustness guarantees against adversarial attacks. In this work, we introduce a framework to obtain a provable adversarial-free region in the neighborhood of the input data by a polyhedr ...

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC2023