---
title: Sparse Autoencoders as Implicit Diversification Regularisers for Cross-Sectional Cryptocurrency Returns
url: https://www.ml-quant.com/papers/arxiv/2610.39205/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-10-30
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2610.39205
source_url: https://arxiv.org/abs/2610.39205
featured: 2026-10-30
citations: unknown
topic: Portfolio & Allocation
---


# Sparse Autoencoders as Implicit Diversification Regularisers for Cross-Sectional Cryptocurrency Returns

Sparse autoencoders extract interpretable factors from cryptocurrency returns and produce lower-volatility portfolios through better asset diversification, though without outperforming simple contrarian strategies.

- Source: https://arxiv.org/abs/2610.39205
- Identifier: arXiv:2610.39205
- Released: 2026-10-30
- First featured: Quant Letter No. 137 (2026-10-30): https://www.ml-quant.com/issues/2026-10-30/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Portfolio & Allocation
- Authors: Weijia Han, Lisha Qu

## Abstract (arXiv, CC0)

Factor models are central to asset pricing and portfolio construction, but neural factors are often difficult to interpret, and their effect on portfolio risk is unclear. A sparse autoencoder (SAE) may address both concerns by describing daily returns across many assets with only a few active features. We ask whether this sparse representation produces stable financial factors and better-diversified portfolios. We study cryptocurrency returns, use equities as a secondary test, and compare the SAE with common benchmarks in a time-ordered evaluation. We find consistently lower volatility in SAE portfolios and link this result to lower correlations among the selected assets. However, the SAE does not outperform a simple contrarian strategy on risk-adjusted return, and its factor identities are not stable across training runs. Our contribution is to show that sparse coding can support portfolio diversification, while clearly separating this risk effect from claims about alpha or stable factor identities.

## Related

- [On Unified Adaptive Black-Litterman Mean-Variance Portfolio Management](https://www.ml-quant.com/papers/arxiv/2307.03391/): The paper presents a new adaptive portfolio management framework that merges dynamic Black-Litterman optimization with the general factor model and Elastic Net regression, showing computational benefits and promising trading results.
- [From Regimes to Portfolios: An Interpretable Machine-Learning Framework for Strategic and Tactical Cross-Asset Allocation](https://www.ml-quant.com/papers/ssrn/7595598/): An interpretable machine-learning framework linking regime identification to cross-asset portfolio construction earns 0.60 net Sharpe, not significantly different from a 60/40 benchmark.
- [Dynamic Portfolio Selection with Factors](https://www.ml-quant.com/papers/repec/eee-dyncon-v-167-y-2024-i-c-s0165188924001155/): A new system of factor models, which considers both return and risk, has been introduced and has shown superior performance in predicting future investment prospects compared to standard policies.
- [XAI in Finance Review](https://www.ml-quant.com/papers/repec/spr-manrev-v-74-y-2024-i-2-d-10-1007-s11301-023-00320-0/): A comprehensive review of 2022 papers on Explainable AI in Finance shows extensive research in risk management and portfolio optimization, but a lack of study in anti-money laundering.
- [An End-to-End Direct Reinforcement Learning Approach for Multi-Factor Based Portfolio Management](https://www.ml-quant.com/papers/ssrn/4729683/): A new online portfolio decision model combines the multifactor model and mean-variance portfolio optimization in one step, enhancing overall performance.
- [Course 2023-2024 in Portfolio Allocation and Asset Management](https://www.ml-quant.com/papers/ssrn/4698165/): The University of Paris-Saclay offers an advanced asset management course covering portfolio optimization, smart beta factor investing, and the use of machine learning in asset management.