A Fully Analog Pipeline for Portfolio Optimization

TL;DR

This paper introduces an energy-efficient, fully analog hardware pipeline that rapidly computes optimal investment portfolios by leveraging physics principles and analog neural networks, reducing computational energy and time.

Contribution

It presents a novel analog approach combining autoencoders and Hopfield networks for portfolio optimization, improving speed and energy efficiency over digital methods.

Findings

Achieves fast portfolio optimization with low energy consumption.

Successfully computes the entire efficient frontier.

Demonstrates accurate covariance matrix estimation using analog autoencoders.

Abstract

Portfolio optimization is a ubiquitous problem in financial mathematics that relies on accurate estimates of covariance matrices for asset returns. However, estimates of pairwise covariance could be better and calculating time-sensitive optimal portfolios is energy-intensive for digital computers. We present an energy-efficient, fast, and fully analog pipeline for solving portfolio optimization problems that overcomes these limitations. The analog paradigm leverages the fundamental principles of physics to recover accurate optimal portfolios in a two-step process. Firstly, we utilize equilibrium propagation, an analog alternative to backpropagation, to train linear autoencoder neural networks to calculate low-rank covariance matrices. Then, analog continuous Hopfield networks output the minimum variance portfolio for a given desired expected return. The entire efficient frontier may then be recovered, and an optimal portfolio selected based on risk appetite.