SciPost Submission Page
Supervised learning of few dirty bosons with variable particle number
by Pere Mujal, Àlex Martínez Miguel, Artur Polls, Bruno Juliá-Díaz, Sebastiano Pilati
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | Pere Mujal · Sebastiano Pilati |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2010.03875v3 (pdf) |
| Data repository: | https://doi.org/10.5281/zenodo.4058492 |
| Date accepted: | 2021-03-11 |
| Date submitted: | 2021-03-02 13:35 |
| Submitted by: | Mujal, Pere |
| Submitted to: | SciPost Physics |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approaches: | Theoretical, Computational |
Abstract
We investigate the supervised machine learning of few interacting bosons in optical speckle disorder via artificial neural networks. The learning curve shows an approximately universal power-law scaling for different particle numbers and for different interaction strengths. We introduce a network architecture that can be trained and tested on heterogeneous datasets including different particle numbers. This network provides accurate predictions for all system sizes included in the training set and, by design, is suitable to attempt extrapolations to (computationally challenging) larger sizes. Notably, a novel transfer-learning strategy is implemented, whereby the learning of the larger systems is substantially accelerated and made consistently accurate by including in the training set many small-size instances.
List of changes
1. In Table 3, we include additional data for the extrapolation from the system sizes N=1 and N=2, to N=4.
2. In Section 4.3, we discuss the additional data concerning the extrapolations from N=1 and N=2 to N=4. The limitations of the extrapolation procedure are further emphasized, highlighting the need to use several system sizes in the training set. The need of a further investigation on the accuracy of the extrapolation procedure was already stressed.
3. In the "Summary and conclusions" Section 5, the possible inaccuracies of the direct extrapolations are more clearly highlighted, mentioning the improvements obtained when larger system sizes are included in the training set.
Published as SciPost Phys. 10, 073 (2021)