How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model

Chat with Graph Search

Ask any question about EPFL courses, lectures, exercises, research, news, etc. or try the example questions below.

DISCLAIMER: The Graph Chatbot is not programmed to provide explicit or categorical answers to your questions. Rather, it transforms your questions into API requests that are distributed across the various IT services officially administered by EPFL. Its purpose is solely to collect and recommend relevant references to content that you can explore to help you answer your questions.

The Birman-Williams theorem gives a connection between the collection of unstable periodic orbits (UPOs) contained within a chaotic attractor and the topology of that attractor, for three-dimensional systems. In certain cases, the fractal dimension of a ch ...

This paper introduces a novel architecture for Quantum Graph Neural Networks, which is significantly different from previous approaches found in the literature. The proposed approach produces similar outcomes with respect to previous models but with fewer ...

To retrieve Surface Solar Radiation (SSR) from satellite images, a baseline reflectance of the observed ground in clear-sky conditions, also called background reflectance, is necessary to distinguish atmospheric absorption and scattering effects from surfa ...

Fuel performance codes are essential for predicting fuel rod behavior in nuclear reactors. These codes comprise a complex system of models with empirical constants that require calibration against experimental data. To support this calibration, this study ...

Monitoring of cardiac output (CO) is a mainstay of hemodynamic management in the acutely or critically ill patient. Invasive determination of CO using thermodilution, albeit regarded as the gold standard, is cumbersome and bears risks inherent to catheteri ...

This study integrates deep learning and geospatial analysis to enhance soil loss estimation in the Moulouya Watershed, a region prone to erosion due to diverse topography and climatic conditions. Traditional models like the Universal Soil Loss Equation (US ...

The digitization of 3D deformable objects remains a significant challenge in computer graphics and vision, particularly in the accurate modeling of garments. Garments exhibit complex shape variability, non-rigid deformations, and frequent self-occlusion, m ...

This study presents an augmented hybrid approach for improving the diagnosis of malignant skin lesions by combining convolutional neural network (CNN) predictions with selective human interventions based on prediction confidence. The algorithm retains high ...

Estimating ecosystem-atmosphere fluxes such as evapotranspiration (ET) in a robust manner and at a global scale remains a challenge. Methods based on machine learning (ML) have shown promising results in achieving such upscaling, providing a complementary ...

Scientific Machine Learning (SciML) is a recently emerged research field which combines physics-based and data-driven models for the numerical approximation of differential problems. Physics-based models rely on the physical understanding of the problem at ...

A robust anomaly detection mechanism should possess the capability to effectively remediate anomalies, restoring them to a healthy state, while preserving essential healthy information. Despite the efficacy of existing generative models in learning the und ...

Object detection, a fundamental task in computer vision, is crucial for various intelligent edge computing applications. However, object detection algorithms are usually heavy in computation, hindering their deployments on resource-constrained edge devices ...

Faithful human performance capture and free-view rendering from sparse RGB observations is a long-standing problem in Vision and Graphics. The main challenges are the lack of observations and the inherent ambiguities of the setting, e.g. occlusions and dep ...

Automatic speech recognition (ASR) systems are well known to perform poorly on dysarthric speech. Previous works have addressed this by speaking rate modification to reduce the mismatch with typical speech. Unfortunately, these approaches rely on transcrib ...

Subjective interpretation and content diversity make predicting whether an image is private or public a challenging task. Graph neural networks combined with convolutional neural networks (CNNs), which consist of 14,000 to 500 millions parameters, generate ...

Machine learning is increasingly integrated with decision making by powering automated systems and informing human decision makers in applications such as CV screening, portfolio allocation, loan risk assessment, and insurance approval. In this thesis, I e ...

As the demand for high channel counts and high-resolution recordings of neural activity continues to grow, the increased power and data rate generated impose hard constraints on the telemetry capabilities of wireless implantable neural interfaces. To addre ...

Image steganalysis, detecting hidden data in digital images, is essential for enhancing digital security. Traditional steganalysis methods typically rely on large, pre-labeled image datasets, which are difficult and costly to compile. To address this, this ...

Small scale ultrasonic piezoelectric actuators performance strongly depends on not well-known contact dynamics. Deep Neural Network (DNN) sees their use in physic simulation growing as their flexibility allows better performance especially when dynamics la ...

How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model

Chat with Graph Search

The topology of a chaotic attractor in the Kuramoto-Sivashinsky equation

A Study on Quantum Graph Neural Networks Applied to Molecular Physics

Retrieval of Surface Solar Radiation through Implicit Albedo Recovery from Temporal Context

Preliminary Sensitivity Analysis of Time-Dependent Code Outputs in Fuel Performance Models Using OFFBEAT for Future Calibration

Machine learning-enabled estimation of cardiac output from peripheral waveforms is independent of blood pressure measurement location in an in silico population

Application of Deep Learning and Geospatial Analysis in Soil Loss Risk in the Moulouya Watershed, Morocco

Neural Garment Modeling

Enhancing skin lesion classification: a CNN approach with human baseline comparison

On the added value of sequential deep learning for the upscaling of evapotranspiration

Combining physics-based and data-driven models: advancing the frontiers of research with scientific machine learning

Self-Supervised Anomaly Segmentation via Diffusion Models with Dynamic Transformer UNet

Edge-Cloud Collaborated Object Detection via Bandwidth Adaptive Difficult-Case Discriminator

MetaCap: Meta-learning Priors from Multi-view Imagery for Sparse-View Human Performance Capture and Rendering

Unsupervised Rhythm and Voice Conversion of Dysarthric to Healthy Speech for ASR

Learning Privacy from Visual Entities

Machine Learning in Decision-Making Systems: Fairness, Robustness, and Data Bias

A Reconfigurable High-Dynamic Range ∆Σ Front-End with Event-Based Decimation for Bandwidth-Efficient Implantable Neural Interfaces

Image steganalysis using active learning and hyperparameter optimization

Normal Contact Force Estimation Using Deep Learning

Replica method for computational problems with randomness: principles and illustrations