• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

RoboScientist: Variational Autoencoder for Automatic Discovery of Laws of Nature

Student: Semavina Iuliia

Supervisor: Andrey Ustyuzhanin

Faculty: Faculty of Computer Science

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Year of Graduation: 2021

When encountering a previously unknown environment, researchers strive to explain its behaviour. Often it can be done with mathematical equations relating some target parameter of the mysterious environment to a number of dependent variables. In this case, the discovery process consists of collecting experimental data from the environment and searching for the equation that fits this data. In this work, we try to automate the discovery process, which we split into two parts. The first part is to recover a formula that fits or at least approximates the given dataset. There exist several approaches to this problem, but each of them has its limitations. For this reason, to tackle this problem, we propose a novel deep learning method based on a recurrent variational autoencoder. We provide the results of applying our algorithm to multiple formulas and show that our method outperforms the baseline on some of them. The second part involves identifying the most informative observations that can simplify further exploration. Although this side of the problem is crucial because the properly chosen dataset might save a lot of time and effort, especially when the experiments require complex models and costly resources, to the best of our knowledge, this part of the problem has not been investigated before. In this work, we propose a method for picking new points for the dataset. We conduct experiments that show the viability of the proposed approach.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses