Paper: arXiv 2602.07659

Authors: Matthew Siper, Muhammad Umair Nasir, Ahmed Khalifa, Lisa Soros, Jay Azhang, Julian Togelius

Abstract

Genetic Programming yields interpretable programs, but small syntactic mutations can induce large, unpredictable behavioral shifts, degrading locality and sample efficiency. We frame this as an operator-design problem: learn a continuous program space where latent distance has behavioral meaning, then design mutation operators that exploit this structure without changing the evolutionary optimizer. We make locality measurable by tracking action-level divergence under controlled latent perturbations, identifying an empirical trust region for behavior-local continuous variation. Using a compact trading-strategy DSL with four semantic components (long/short entry and exit), we learn a matching block-factorized embedding and compare isotropic Gaussian mutation over the full latent space to geometry-compiled mutation that restricts updates to semantically paired entry–exit subspaces and proposes directions using a learned flow-based model trained on logged mutation outcomes. Under identical $(μ+λ)$ evolution strategies and fixed evaluation budgets across five assets, the learned mutation operator discovers strong strategies using an order of magnitude fewer evaluations and achieves the highest median out-of-sample Sharpe ratio. Although isotropic mutation occasionally attains higher peak performance, geometry-compiled mutation yields faster, more reliable progress, demonstrating that semantically aligned mutation can substantially improve search efficiency without modifying the underlying evolutionary algorithm.

Complexity vs Empirical Score

  • Math Complexity: 5.0/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a novel approach to Genetic Programming in quantitative trading by embedding programs into a factorized continuous latent space and using a flow-based model for geometry-compiled mutations. Empirical rigor is strong, demonstrating a 7.3x improvement in sample efficiency and higher out-of-sample Sharpe ratios across multiple assets compared to standard baselines.

Research Flowchart

  flowchart TD
    A["Research Goal: Improve Genetic Programming locality and efficiency via a learned continuous program space"] --> B["Inputs: Compact Trading DSL & 5 Asset Data Streams"]
    B --> C["Encoder: Learn Block-Factorized Embedding aligned with Entry/Exit Semantics"]
    C --> D["Quantify Locality: Measure Action Divergence to Define Behavioral Trust Region"]
    D --> E["Search Process: (μ+λ) Evolution Strategy comparing Isotropic Gaussian vs. Geometry-Compiled Flow Mutation"]
    E --> F["Key Outcomes: 10x Fewer Evaluations & Highest Median Out-of-Sample Sharpe Ratio"]

Cited References (28)

  1. Genetic Programming: On the Programming of Computers by Means of Natural Selection (1992)
  2. Deepmasterprints: Generating masterprints for dictionary attacks via latent variable evolution (2018)
  3. Genetic Programming II: Automatic Discovery of Reusable Programs (1994)
  4. Mutation models: Learning to generate levels by imitating evolution (2022)
  5. Grammatically-based Genetic Programming (1995)
  6. A preliminary study of disentanglement with insights on the inadequacy of metrics (2019) — link
  7. Strongly Typed Genetic Programming (1995)
  8. Geometric Semantic Genetic Programming (2012)
and 20 more
  1. Semantic Backpropagation for Designing Search Operators in Genetic Programming (2014)
  2. Neuroevolution: From Architectures to Learning (2008)
  3. Designing Neural Networks through Neuroevolution (2019)
  4. Completely Derandomized Self-Adaptation in Evolution Strategies (2001)
  5. Evolution Strategies as a Scalable Alternative to Reinforcement Learning (2017) — link
  6. Auto-Encoding Variational Bayes (2014) — link
  7. Program Synthesis as Latent Continuous Optimization: Evolutionary Search in Neural Embeddings (2020) — link
  8. Learnable Embeddings of Program Spaces (2011)
  9. Program Synthesis in a Continuous Space Using Grammars and Variational Autoencoders (2020) — link
  10. Evolving through the Looking Glass: Learning Improved Search Spaces with Variational Autoencoders (2022)
  11. Representation Learning: A Review and New Perspectives (2013)
  12. beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework (2017)
  13. Disentangling by Factorising (2018)
  14. Isolating Sources of Disentanglement in Variational Autoencoders (2018)
  15. A Framework for the Quantitative Evaluation of Disentangled Representations (2018)
  16. Probabilistic Model Building in Genetic Programming: A Critical Review (2014) — link
  17. Denoising Autoencoder Genetic Programming: Strategies to Control Exploration and Exploitation in Search (2023)
  18. Denoising Diffusion Probabilistic Models (2020)
  19. Denoising Diffusion Implicit Models (2021)
  20. Riccardo Poli, William B. Langdon, Nicholas F. McPhee: A Field Guide to Genetic Programming: Lulu. com, 2008, 250 pp, ISBN 978-1-4092-0073-4 (2009)

0 of 28 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.