Vision Transformer Adapters for Generalizable Multitask Learning

Publications associées (23)

À propos
Confidentialité
Mentions légales

Graph Chatbot

Fast Transformers with Clustered Attention

François Fleuret, Angelos Katharopoulos, Apoorv Vyas

Transformers have been proven a successful model for a variety of tasks in sequence modeling. However, computing the attention matrix, which is their key component, has quadratic complexity with respect to the sequence length, thus making them prohibitivel ...

2020

Visual Correspondences for Unsupervised Domain Adaptation on Electron Microscopy Images

Pascal Fua, Mathieu Salzmann, Carlos Joaquin Becker, Róger Bermúdez Chacón, Okan Altingövde

We present an Unsupervised Domain Adaptation strategy to compensate for domain shifts on Electron Microscopy volumes. Our method aggregates visual correspondences—motifs that are visually similar across different acquisitions—to infer changes on the parame ...

Institute of Electrical and Electronics Engineers2020

Context-Aware Attention Mechanism for Speech Emotion Recognition

Philip Neil Garner, Gaétan Ramet

In this work, we study the use of attention mechanisms to enhance the performance of the state-of-the-art deep learning model in Speech Emotion Recognition (SER). We introduce a new Long Short-Term Memory (LSTM)-based neural network attention model which i ...

IEEE2018

Program Analysis and Compilation Techniques for Speeding up Transactional Database Workloads

Mohammad Dashti Rahmat Abadi

There is a trend towards increased specialization of data management software for performance reasons. The improved performance not only leads to a more efficient usage of the underlying hardware and cuts the operation costs of the system, but also is a ga ...

EPFL2017

Sequential Proximity: Towards Provably Scalable Concurrent Search Algorithms

Rachid Guerraoui, Vasileios Trigonakis, Karolos Antoniadis, Julien Michel Stainer

Establishing the scalability of a concurrent algorithm a priori, before implementing and evaluating it on a concrete multi-core platform, seems difficult, if not impossible. In the context of search data structures however, according to all practical work ...

Springer International Publishing2017

Universally Scalable Concurrent Data Structures

Tudor Alexandru David

The increase in the number of cores in processors has been an important trend over the past decade. In order to be able to efficiently use such architectures, modern software must be scalable: performance should increase proportionally to the number of all ...

EPFL2017

Transactions Chasing Scalability and Instruction Locality on Multicores

Pinar Tözün

For several decades, online transaction processing (OLTP) has been one of the main server applications that drives innovations in the data management ecosystem, and in turn the database and computer architecture communities. Recent hardware trends oblige s ...

EPFL2014

Scaling up Mixed Workloads: a Battle of Data Freshness, Flexibility, and Scheduling

Anastasia Ailamaki, Iraklis Psaroudakis

The common "one size does not fit all" paradigm isolates transactional and analytical workloads into separate, specialized database systems. Operational data is periodically replicated to a data warehouse for analytics. Competitiveness of enterprises today ...

Springer International Publishing AG2015

Scaling up analytical queries with column-stores

Anastasia Ailamaki, Ioannis Alagiannis, Manoussos Gavriil Athanassoulis

As data analytics is used by an increasing number of applications, data analytics engines are required to execute workloads with increased concurrency, i.e., an increasing number of clients submitting queries. Data management systems designed for data anal ...

2013

Sharing Data and Work Across Concurrent Analytical Queries

Anastasia Ailamaki, Manoussos Gavriil Athanassoulis, Iraklis Psaroudakis

Today's data deluge enables organizations to collect massive data, and analyze it with an ever-increasing number of concurrent queries. Traditional data warehouses (DW) face a challenging problem in executing this task, due to their query-centric model: ea ...

2013

Page 2 sur 3