Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GPU-Accelerated Rule Evaluation and Evolution

Hormoz Shahrzad, Risto Miikkulainen

TL;DR

Evolutionary Rule-based Learning (ERL) methods suffer from expensive rule evaluation on large datasets. The authors propose Accelerated Evolutionary Rule-based Learning (AERL), which uses a tensorized rule representation in PyTorch to enable GPU-based parallel evaluation and backpropagation-based fine-tuning of rule coefficients. This approach encodes rules as inequalities in a tensor form, allowing matrix operations for evaluation while preserving logical structure and enabling gradient-based refinement. Experiments on the UCI Breast Cancer dataset demonstrate substantial GPU speedups over CPU evaluation and faster evolutionary progress, highlighting improved scalability and potential for explainable AI.

Abstract

This paper introduces an innovative approach to boost the efficiency and scalability of Evolutionary Rule-based machine Learning (ERL), a key technique in explainable AI. While traditional ERL systems can distribute processes across multiple CPUs, fitness evaluation of candidate rules is a bottleneck, especially with large datasets. The method proposed in this paper, AERL (Accelerated ERL) solves this problem in two ways. First, by adopting GPU-optimized rule sets through a tensorized representation within the PyTorch framework, AERL mitigates the bottleneck and accelerates fitness evaluation significantly. Second, AERL takes further advantage of the GPUs by fine-tuning the rule coefficients via back-propagation, thereby improving search space exploration. Experimental evidence confirms that AERL search is faster and more effective, thus empowering explainable artificial intelligence.

GPU-Accelerated Rule Evaluation and Evolution

TL;DR

Evolutionary Rule-based Learning (ERL) methods suffer from expensive rule evaluation on large datasets. The authors propose Accelerated Evolutionary Rule-based Learning (AERL), which uses a tensorized rule representation in PyTorch to enable GPU-based parallel evaluation and backpropagation-based fine-tuning of rule coefficients. This approach encodes rules as inequalities in a tensor form, allowing matrix operations for evaluation while preserving logical structure and enabling gradient-based refinement. Experiments on the UCI Breast Cancer dataset demonstrate substantial GPU speedups over CPU evaluation and faster evolutionary progress, highlighting improved scalability and potential for explainable AI.

Abstract

This paper introduces an innovative approach to boost the efficiency and scalability of Evolutionary Rule-based machine Learning (ERL), a key technique in explainable AI. While traditional ERL systems can distribute processes across multiple CPUs, fitness evaluation of candidate rules is a bottleneck, especially with large datasets. The method proposed in this paper, AERL (Accelerated ERL) solves this problem in two ways. First, by adopting GPU-optimized rule sets through a tensorized representation within the PyTorch framework, AERL mitigates the bottleneck and accelerates fitness evaluation significantly. Second, AERL takes further advantage of the GPUs by fine-tuning the rule coefficients via back-propagation, thereby improving search space exploration. Experimental evidence confirms that AERL search is faster and more effective, thus empowering explainable artificial intelligence.
Paper Structure (8 sections, 3 figures)

This paper contains 8 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Method
  4. Experimental Results
  5. Experiment Setup
  6. Evolutionary Performance
  7. Computational Efficiency
  8. Discussion and Future Work

Figures (3)

  • Figure 1: Example of rule set representation and fine-tuning. A randomly generated rule set (not evolved) is shown in both numerical (tensor) form and human-readable form to illustrate how feature weights and biases translate into logical conditions. Back-propagation is applied to tune the coefficients, demonstrating how fine-tuning alone can substantially improve accuracy while keeping the logical structure intact.
  • Figure 2: Evolutionary Progress of PyTorch Rules in the UCI Breast Cancer Domain with vs. without Fine-tuning. Each point is averaged over 10 runs, with 95% confidence intervals. In both cases, rule structures are evolved, but in the fine-tuning setting, back-propagation is applied at each generation to optimize rule coefficients without altering structure. This gradient-based refinement improves accuracy and accelerates search via the Baldwin effect ($p = 2 \times 10^{-7}$).
  • Figure 3: Comparison of GPU to CPU Acceleration Ratios with and without Fine-tuning. Plots show average CPU/GPU wallclock time ratio over 10 runs with 95% confidence intervals. Both with and without fine-tuning, GPU acceleration provides a minimum fivefold speedup compared to CPUs. Observed variations are due to the total number of rules and conditions per generation within a fixed pool size of 100 candidates. These results demonstrate that GPUs significantly accelerate rule-set evolution with and without fine-tuning.