Publications associées à MP3

Communication-efficient distributed training of machine learning models

In this thesis, we explore techniques for addressing the communication bottleneck in data-parallel distributed training of deep learning models. We investigate algorithms that either reduce the size of the messages that are exchanged between workers, or th ...

EPFL2023

Large-scale traffic signal control and multimodal network design

Dimitrios Tsitsokas

Traffic congestion constitutes one of the most frequent, yet challenging, problems to address in the urban space. Caused by the concentration of population, whose mobility needs surpass the serving capacity of urban networks, congestion cannot be resolved ...

EPFL2022

Volume change response and fabric evolution of granular MX80 bentonite along different hydro-mechanical stress paths

Alessio Ferrari, Jose Antonio Bosch Llufriu, Patrycja Ewelina Baryla

Despite the increasing understanding of bentonite behaviour, there is still missing evidence on how different hydro-mechanical loadings, including sequences of hydration and compression, affect the fabric and the volume change behaviour of the material. It ...

Springer Nature2022

Systems and methods for digital information decoding and data storage in hybrid macromolecules

Aleksandra Radenovic, Matteo Dal Peraro, Chan Cao, Lucien Fabrice Krapp

Systems and methods for data storage and readout using hybrid nucleic acid- polymeric molecules are herein disclosed. A molecular data storage medium is presented comprising a header and a footer, each comprising or consisting of at least one unit of a fir ...

2022

Learnable filter-banks for CNN-based audio applications

Nicolas Aspert, Benjamin Ricaud, Helena Peic Tukuljac, Laurent Colbois

We investigate the design of a convolutional layer where kernels are parameterized functions. This layer aims at being the input layer of convolutional neural networks for audio applications or applications involving time-series. The kernels are defined as ...

2022

Two-layer adaptive signal control framework for large-scale networks combining efficient Max-Pressure and Perimeter Control

Nikolaos Geroliminis, Dimitrios Tsitsokas, Anastasios Kouvelas

This work investigates the benefits of a two-layer adaptive signal control framework combining multi-region perimeter control (PC) with distributed Max Pressure (MP) control in selected network intersections. Motivated by MP’s questionable performance in o ...

2022

Benchmarking of objective quality metrics for point cloud compression

Touradj Ebrahimi, Davi Nachtigall Lazzarotto, Evangelos Alexiou

Point cloud is a promising imaging modality for the representation of 3D media. The vast volume of data associated with it requires efficient compression solutions, with lossy algorithms leading to larger bit-rate savings at the expense of visual impairmen ...

IEEE2021

Quality Evaluation Of Static Point Clouds Encoded Using MPEG Codecs

Touradj Ebrahimi, Evangelos Alexiou

This paper presents a quality evaluation study of point cloud codecs that have been recently standardised by the MPEG committee. In particular, a subjective experiment to assess their performance in terms of bitrate against visual quality is designed and r ...

IEEE2020

Subjective and Viewport-based Objective Quality Assessment of 360-degree Videos

Pascal Frossard, Roberto Gerson De Albuquerque Azevedo

Visual distortions in processed 360-degree visual content and consumed through head-mounted displays (HMDs) are perceived very differently when compared to traditional 2D content. To better understand how compression-related artifacts affect the overall pe ...

2020

Fast Block Sequential Decoding of Polar Codes

Kirill Ivanov

A reduced complexity sequential decoding algorithm for polar (sub) codes is described. The proposed approach relies on a decomposition of the polar (sub) code being decoded into a number of outer codes, and on-demand construction of codewords of these code ...

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC2020

Practical Low-Rank Communication Compression in Decentralized Deep Learning

Martin Jaggi, Sai Praneeth Reddy Karimireddy, Thijs Vogels

Lossy gradient compression has become a practical tool to overcome the communication bottleneck in centrally coordinated distributed training of machine learning models. However, algorithms for decentralized training with compressed communication over arbi ...

2020

A comprehensive study of the rate-distortion performance in MPEG point cloud compression

Touradj Ebrahimi, Evangelos Alexiou, Irene Viola

Recent trends in multimedia technologies indicate the need for richer imaging modalities to increase user engagement with the content. Among other alternatives, point clouds denote a viable solution that offers an immersive content representation, as witne ...

2019

On the relevance of quality score metadata in genomic sequence data for omics applications

Ana Angelica Hernandez Lopez

High-throughput sequencing of DNA molecules has revolutionized biomedical research by enabling the quantitative analysis of the genome to study its function, structure and dynamics. It is driving sequencing-based experiments in life sciences as evidenced b ...

EPFL2019

Towards an Algebraic Network Information Theory: Distributed Lossy Computation of Linear Functions

Michael Christoph Gastpar, Sung Hoon Lim, Adriano Pastore, Chen Feng

Consider the important special case of the K-user distributed source coding problem where the decoder only wishes to recover one or more linear combinations of the sources. The work of Körner and Marton demonstrated that, in some cases, the optimal rate re ...

2019

Gaussian channel transmission of images and audio files using cryptcoding

Vladimir Ilievski

Random codes based on quasigroups (RCBQ) are cryptcodes, i.e. they are error-correcting codes, which provide information security. Cut-Decoding and 4-Sets-Cut-Decoding algorithms for these codes are defined elsewhere. Also, the performance of these codes f ...

2019

3D Box Proposals from a Single Monocular Image of an Indoor Scene

Mathieu Salzmann, Wei Zhuo

Modern object detection methods typically rely on bounding box proposals as input. While initially popularized in the 2D case, this idea has received increasing attention for 3D bounding boxes. Nevertheless, existing 3D box proposal techniques all assume h ...

ASSOC ADVANCEMENT ARTIFICIAL INTELLIGENCE2018

Approximate Cloaking via Transformation Optics for Electromagnetic Waves

Xuan Loc Tran

Cloaking via transformation optics was introduced by Pendry, Schurig, and Smith for the Maxwell system and Leonhardt in the geometric optics setting. They used a singular change of variables which blows up a point into a cloaked region. The same transforma ...

EPFL2018

Adaptive Learning-Based Compressive Sampling for Low-power Wireless Implants

Volkan Cevher, Catherine Dehollain, Mahsa Shoaran, Yusuf Leblebici, Gürkan Yilmaz, Cosimo Aprile, Luca Baldassarre, Kerim Türe, Franco Maloberti

Implantable systems are nowadays being used to interface the human brain with external devices, in order to understand and potentially treat neurological disorders. The most predominant design constraints are the system’s area and power. In this paper, we ...

2018

Deep Learning of Human Perception in Audio Event Classification

Samuel Denys Beuret, Yi Yu

In this paper, we introduce our recent studies on human perception in audio event classification. In particular, the pre-trained model VGGish is used as feature extractor to process audio data, and DenseNet is trained by and used as feature extractor for o ...

IEEE2018

Quality assessment of an HDR dual-layer backward-compatible codec compared to uncompromised SDR and HDR solutions

Touradj Ebrahimi, Martin Rerabek, Anne-Flore Nicole Marie Perrin

Broadcasting high dynamic range (HDR) video has been demonstrated as largely preferred when compared to standard dynamic range (SDR), mainly due to its capability of representing more details in dark and bright regions. Additionally, progress over the last ...

2018