278 citations · 792 across the 28 of their papers we have counts for
7 papers · 1 filter
CaReAQA: A Cardiac and Respiratory Audio Question Answering Model for Open-Ended Diagnostic Reasoning
Tsai-Ning Wang, Lin-Lin Chen, Neil Zeghidour +1
Medical audio signals, such as heart and lung sounds, play a crucial role in clinical diagnosis. However, analyzing these signals remains challenging: traditional methods rely on h…
MusicRL: Aligning Music Generation to Human Preferences
Geoffrey Cideron, Sertan Girgin, Mauro Verzetti +11
We propose MusicRL, the first music generation system finetuned from human feedback. Appreciation of text-to-music models is particularly subjective since the concept of musicality…
DNArch: Learning Convolutional Neural Architectures by Backpropagation
David W. Romero, Neil Zeghidour
We present Differentiable Neural Architectures (DNArch), a method that jointly learns the weights and the architecture of Convolutional Neural Networks (CNNs) by backpropagation. I…
Learning strides in convolutional neural networks
Rachid Riad, Olivier Teboul, David Grangier +1
Convolutional neural networks typically contain several downsampling operators, such as strided convolutions or pooling layers, that progressively reduce the resolution of intermed…
General-purpose, long-context autoregressive modeling with Perceiver AR
Curtis Hawthorne, Andrew Jaegle, Cătălina Cangea +12
Real-world data is high-dimensional: a book, image, or musical performance can easily contain hundreds of thousands of elements even after compression. However, the most commonly u…
Deep multi-class learning from label proportions
Gabriel Dulac-Arnold, Neil Zeghidour, Marco Cuturi +2
We propose a learning algorithm capable of learning from label proportions instead of direct data labels. In this scenario, our data are arranged into various bags of a certain siz…
To Reverse the Gradient or Not: An Empirical Comparison of Adversarial and Multi-task Learning in Speech Recognition
Yossi Adi, Neil Zeghidour, Ronan Collobert +3
Transcribed datasets typically contain speaker identity for each instance in the data. We investigate two ways to incorporate this information during training: Multi-Task Learning…