We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

q-fin.PM

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Quantitative Finance > Portfolio Management

Title: Dirichlet policies for reinforced factor portfolios

Abstract: This article aims to combine factor investing and reinforcement learning (RL). The agent learns through sequential random allocations which rely on firms' characteristics. Using Dirichlet distributions as the driving policy, we derive closed forms for the policy gradients and analytical properties of the performance measure. This enables the implementation of REINFORCE methods, which we perform on a large dataset of US equities. Across a large range of implementation choices, our result indicates that RL-based portfolios are very close to the equally-weighted (1/N) allocation. This implies that the agent learns to be agnostic with regard to factors. This is partly consistent with cross-sectional regressions showing a strong time variation in the relationship between returns and firm characteristics.
Comments: 37 pages
Subjects: Portfolio Management (q-fin.PM); Mathematical Finance (q-fin.MF); Machine Learning (stat.ML)
MSC classes: 60J05, 91G10, 91G60
ACM classes: G.3
Cite as: arXiv:2011.05381 [q-fin.PM]
  (or arXiv:2011.05381v2 [q-fin.PM] for this version)

Submission history

From: Guillaume Coqueret [view email]
[v1] Tue, 10 Nov 2020 20:25:41 GMT (702kb,D)
[v2] Thu, 12 Nov 2020 08:25:15 GMT (705kb,D)

Link back to: arXiv, form interface, contact.