Periodic extrapolative generalisation in neural networks

P Belcak, R Wattenhofer - 2022 IEEE Symposium Series on …, 2022 - ieeexplore.ieee.org
The learning of the simplest possible computational pattern-periodicity-is an open problem
in the research of strong generalisation in neural networks. We formalise the problem of …

Neural combinatorial logic circuit synthesis from input-output examples

P Belcak, R Wattenhofer - arxiv preprint arxiv:2210.16606, 2022 - arxiv.org
We propose a novel, fully explainable neural approach to synthesis of combinatorial logic
circuits from input-output examples. The carrying advantage of our method is that it readily …

Towards Foundation Models with Mathematical Understanding

P Belcak, R Wattenhofer - ICLR 2023 Workshop on Mathematical …, 2023 - openreview.net
We investigate the ability of transformer models to build representations of integer
sequences that are of utility to tasks where deeper mathematical understanding is needed …

[PDF][PDF] Overview of Achievements

R Wattenhofer - 2024 - disco.ethz.ch
Objective Overview of Achievements Page 1 Roger Wattenhofer ETH Zurich, 8092 Zurich,
Switzerland phone +41 44 632 6312, fax +41 44 632 1035 wattenhofer@ethz.ch, www.disco.ethz.ch …