TL;DR —
State-of-the-art machine learning models struggle with generalization which can only be achieved by proper accounting for core knowledge priors.
This paper is available on arxiv under CC 4.0 license.
Authors:
(1) Mattia Atzeni, EPFL, Switzerland and mattia.atzeni@outlook.it;
(2) Mrinmaya Sachan, ETH Zurich, Switzerland;
(3) Andreas Loukas, Prescient Design, Switzerland.
Table of Links
- Abstract & Introduction
- Formalizing the Group-Action Learning Problem
- Attention Masks for Core Geometry Priors
- The LATFORMER Architecture
- Experiments
- Related Work
- Conclusion & References
- A. Additional Details on the Model
- B. Additional Experiments and Details on the Experimental Setup
- C. Limitations and Future Work
- D. Deferred Proofs
D. Deferred Proofs
We prove both Theorem 3.1 and 3.2 by induction on the dimensionality of the hypercubic lattice m.
D.1. Base Case for Theorems 1 and 2
D.2. Inductive Step for Theorems 1 and 2
D.3. Proof of Corollary 1
The proof of Corollary 1 follows immediately from Theorem 3.2 and from the property of the Fourier transform according to which multiplying in the Fourier domain implements a convolution in the original domain.
[story continues]
Written by
@escholar
We publish the best academic work (that's too often lost to peer reviews & the TA's desk) to the global tech community
Topics and
tags
tags
abstraction-corpus|reasoning-corpus|arc|languagecomplete-instantiation|larc|non-learning-ai|latformer|neural-network-learning
This story on HackerNoon has a decentralized backup on Sia.
Transaction ID: Q1nHaN5plEO2vyfYlBljvUXcIcWBp4yJFfeUb4Ot36o