Zobrazeno 1 - 10
of 127
pro vyhledávání: '"Rodola, Emanuele"'
Autor:
Zhou, Luca, Solombrino, Daniele, Crisostomi, Donato, Bucarelli, Maria Sofia, Silvestri, Fabrizio, Rodolà, Emanuele
Model merging has recently emerged as a cost-efficient paradigm for multi-task learning. Among current approaches, task arithmetic stands out for its simplicity and effectiveness. In this paper, we motivate the effectiveness of task vectors by linkin
Externí odkaz:
http://arxiv.org/abs/2411.03055
Autor:
Basile, Lorenzo, Maiorca, Valentino, Bortolussi, Luca, Rodolà, Emanuele, Locatello, Francesco
When examined through the lens of their residual streams, a puzzling property emerges in transformer networks: residual contributions (e.g., attention heads) sometimes specialize in specific tasks or input attributes. In this paper, we analyze this p
Externí odkaz:
http://arxiv.org/abs/2411.00246
Deep neural networks often learn similar internal representations, both across different models and within their own layers. While inter-network similarities have enabled techniques such as model stitching and merging, intra-network similarities pres
Externí odkaz:
http://arxiv.org/abs/2410.04941
Autor:
Miranda, Michele, Ruzzetti, Elena Sofia, Santilli, Andrea, Zanzotto, Fabio Massimo, Bratières, Sébastien, Rodolà, Emanuele
Large Language Models (LLMs) represent a significant advancement in artificial intelligence, finding applications across various domains. However, their reliance on massive internet-sourced datasets for training brings notable privacy issues, which a
Externí odkaz:
http://arxiv.org/abs/2408.05212
The emergence of similar representations between independently trained neural models has sparked significant interest in the representation learning community, leading to the development of various methods to obtain communication between latent space
Externí odkaz:
http://arxiv.org/abs/2406.15057
Neural models learn data representations that lie on low-dimensional manifolds, yet modeling the relation between these representational spaces is an ongoing challenge. By integrating spectral geometry principles into neural modeling, we show that th
Externí odkaz:
http://arxiv.org/abs/2406.14183
In this paper, we present a novel data-free method for merging neural networks in weight space. Differently from most existing works, our method optimizes for the permutations of network neurons globally across all layers. This allows us to enforce c
Externí odkaz:
http://arxiv.org/abs/2405.17897
In this work, we present the local patch mesh representation for neural signed distance fields. This technique allows to discretize local regions of the level sets of an input SDF by projecting and deforming flat patch meshes onto the level set surfa
Externí odkaz:
http://arxiv.org/abs/2405.12895
Autor:
Postolache, Emilian, Polouliakh, Natalia, Kitano, Hiroaki, Connelly, Akima, Rodolà, Emanuele, Cosmo, Luca, Akama, Taketo
In this article, we explore the potential of using latent diffusion models, a family of powerful generative models, for the task of reconstructing naturalistic music from electroencephalogram (EEG) recordings. Unlike simpler music with limited timbre
Externí odkaz:
http://arxiv.org/abs/2405.09062
Autor:
Ciranni, Ruben, Mariani, Giorgio, Mancusi, Michele, Postolache, Emilian, Fabbro, Giorgio, Rodolà, Emanuele, Cosmo, Luca
We present COCOLA (Coherence-Oriented Contrastive Learning for Audio), a contrastive learning method for musical audio representations that captures the harmonic and rhythmic coherence between samples. Our method operates at the level of the stems co
Externí odkaz:
http://arxiv.org/abs/2404.16969