From Bricks to Bridges: Product of Invariances to Enhance Latent Space Communication

Abstract

It has been observed that representations learned by distinct neural networks conceal structural similarities when the models are trained under similar inductive biases. From a geometric perspective, identifying the classes of transformations and the related invariances that connect these representations is fundamental to unlocking applications, such as merging, stitching, and reusing different neural modules. However, estimating task-specific transformations a priori can be challenging and expensive due to several factors (e.g., weights initialization, training hyperparameters, or data modality). To this end, we introduce a versatile method to directly incorporate a set of invariances into the representations, constructing a product space of invariant components on top of the latent representations without requiring prior knowledge about the optimal invariance to infuse. We validate our solution on classification and reconstruction tasks, observing consistent latent similarity and downstream performance improvements in a zero-shot stitching setting. The experimental analysis comprises three modalities (vision, text, and graphs), twelve pretrained foundational models, eight benchmarks, and several architectures trained from scratch.

Publication
International Conference on Learning Representations (ICLR 2024)
Irene Cannistraci
Irene Cannistraci
PhD Student

I am a Ph.D. student passionate about Artificial Intelligence.

Luca Moschella
Luca Moschella
PhD Student

PhD Student @SapienzaRoma CS | Intern @NVIDIA Toronto Lab | @NNAISENSE

Marco Fumero
Marco Fumero
PhD Student
Valentino Maiorca
Valentino Maiorca
PhD Student

PhD student @ Sapienza, University of Rome

Emanuele Rodolà
Emanuele Rodolà
Full Professor