---
title: AI Shrinkage for Portfolios
url: https://www.ml-quant.com/papers/ssrn/5261738/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: SSRN 5261738
source_url: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5261738
featured: 2025-05-21
citations: unknown
topic: Portfolio & Allocation
---


# AI Shrinkage for Portfolios

The article introduces a new AI-based tool that enhances the efficiency and performance of risk-optimized portfolios by correcting biases in traditional estimates.

- Source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5261738
- Identifier: SSRN 5261738
- Released: 2025-05-20
- First featured: Quant Letter No. 98 (2025-05-21): https://www.ml-quant.com/issues/2025-05-21/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Portfolio & Allocation

## Related

- [A Unified Framework for Fast Large-Scale Portfolio Optimization](https://www.ml-quant.com/papers/arxiv/2303.12751/): A new large-scale portfolio optimization framework, using shrinkage and regularization techniques, has been tested and proven effective using 50 years of US company return data.
- [Dynamic Realized Minimum Variance Portfolio Models](https://www.ml-quant.com/papers/arxiv/2310.13511/): The article introduces a new model for predicting future portfolios using high-frequency financial data, which minimizes variance using the least absolute shrinkage and selection operator.
- [Covariance matrix filtering and portfolio optimisation: the average oracle vs non-linear shrinkage and all the variants of DCC-NLS](https://www.ml-quant.com/papers/arxiv/2309.17219/): The Average Oracle, a fast covariance filtering method, outperforms complex methods, yielding superior Sharpe ratios in large-scale experiments.
- [Covariance Matrix Shrinkage](https://www.ml-quant.com/papers/repec/eee-ecmode-v-144-y-2025-i-c-s0264999324003389/): The study suggests an optimal shrinkage intensity selection for the linear shrinkage estimator family, which results in more stable covariance matrix estimators and improves global minimum-variance portfolios.
- [Beating Diversification Strategies](https://www.ml-quant.com/papers/repec/cup-jfinqa-v-59-y-2024-i-8-p-3601-3632-3/): The research indicates that the $1/N$ rule is best in high-dimensionality but can be improved by combining it with other rules or machine learning portfolios.
- [Consistent estimation of the high-dimensional efficient frontier](https://www.ml-quant.com/papers/arxiv/2409.15103/): The paper uses random matrix theory to study the asymptotic behavior of the mean-variance efficient frontier, finding that two out of three key quantities are biased and overestimated in high-dimensional asymptotic conditions.