Research internship - Criticality is all you need

ABOUT THE COMPANY

Founded in 1991, we are a global quantitative and systematic asset management firm applying a scientific approach to finance to develop alternative investment strategies that create value for our clients.

We value innovation, dedication, collaboration, and the ability to make an impact. Together, we create a stimulating environment for talented and passionate experts in research, technology, and business to explore new ideas and challenge existing assumptions.

Overview
Recurrent neural networks (RNNs) are extremely popular models for time series learning. A seminal study has showed that chaotic dynamics can emerge sharply in these networks at a critical value of the initialization parameters [1]. Further work revealed that learning is favoured at or even slightly beyond the so-called “edge of chaos” critical point [2, 3]. By exploring the edge of chaos, these networks can enhance their computational capacity and improve predictive performance for time series exhibiting noise and irregular patterns [2].
The aim of this project is to study the predictive capability of RNNs for noisy time-series around their critical point, comparing different architectures and trying to characterize the criticality with concepts borrowed from statistical physics.

Goals
• Review of the current literature on the topic, with emphasis on reservoir computing and applications to time series prediction.
• Characterization of the critical regime and the edge of chaos, its scaling and universal properties, for various RNN architectures.
• Study of the chaotic regime and the related Lyapunov spectrum.
• Development of toy-models and heuristics that can help for training RNNs in the little-data high-noise regime.
This work will be carried out in collaboration with the Short-term Equity Alpha Research Team at CFM, focusing on developing trading strategies and in close cooperation with market data specialists and data scientists.

Profile

Given the research-oriented nature of this project, applicants should have a foundational understanding of statistics, machine learning and have been exposed to the ideas of statistical physics. A good knowledge of python and an understanding of code architecture and data structures will be required. Masters' student/end of study internship preferred.

Duration

6 months preferred -Flexible starting date

References
[1] - H. Sompolinsky, A. Crisanti, and H. J. Sommers, "Chaos in Random Neural Networks", Phys Rev. Lett. 61, 259 - 1988
[2]- Rainer Engelken, Fred Wolf, and L. F. Abbott, “Lyapunov spectra of chaotic recurrent neural networks”, Phys. Rev. Research 5, 043044 (2023).
[3] - Daniel J. Gauthier, Erik Bollt, Aaron Griffith & Wendson A. S. Barbosa, "Next generation reservoir computing", Nature Comm. 12, 5564 (2021).

EQUAL OPPORTUNITIES STATEMENT

We are continuously striving to be an equal opportunity employer and we prohibit any discrimination based on sex, disability, origin, sexual orientation, gender identity, age, race, or religion. We believe that our diversity, breadth of experience, and multiple points of view are among the leading factors in our success.

CFM is a signatory of the Women Empowerment Principles.

FOLLOW US

Follow us on Twitter or LinkedIn or visit our website to find out more about CFM.