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.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the efficiency and scalability of Evolutionary Rule - based Machine Learning (ERL) systems, especially the performance bottleneck problem when dealing with large - scale data sets. Specifically: 1. **Bottlenecks of Traditional ERL Systems**: Although traditional ERL systems can distribute processing tasks across multiple CPUs, there are bottlenecks when evaluating the fitness of candidate rules, especially when dealing with large - scale data sets, and this process becomes particularly time - consuming. 2. **GPU - Accelerated Solutions**: To solve the above problems, this paper proposes the AERL (Accelerated ERL) method, which significantly improves the efficiency of ERL in the following two ways: - **Adopting GPU - Optimized Rule Sets**: By using a tensorial representation in the PyTorch framework, AERL can significantly accelerate fitness evaluation and alleviate the bottleneck problem. - **Using GPU for Fine - Tuning of Rule Coefficients**: By using the back - propagation technique to fine - tune the rule coefficients, it further improves the exploration of the search space, thereby increasing the search speed and effectiveness. 3. **Experimental Verification**: The experimental results show that AERL is not only faster in search speed but also more effective in search results, thereby enhancing the capabilities of explainable artificial intelligence. ### Formulas and Technical Details - **Tensorial Representation of Rule Sets**: Convert logical rules into numerical forms for efficient processing on the GPU. For example, rule conditions are represented as inequalities and multiple rules are simultaneously evaluated through matrix multiplication. The specific representation is as follows: \[ \text{Rule 1: } \begin{bmatrix} 1.0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 1.0 & - 0.38 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1.0 & 0 & 0 \end{bmatrix}, \quad \text{Biases: } \begin{bmatrix} - 2.02 \\ 0.0 \\ - 0.29 \end{bmatrix} \] \[ \text{Rule 2: } \begin{bmatrix} 0 & 0 & 0 & 0 & 1.0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 1.0 \end{bmatrix}, \quad \text{Biases: } \begin{bmatrix} - 0.43 \\ - 0.29 \end{bmatrix} \] - **Application of Activation Functions**: In order to convert continuous output into binary decisions (true or false), AERL applies activation functions. For example, use a clipping function to set negative values to 0 and positive values to 1. For the training or fine - tuning stage, a smooth Sigmoid function can be used to provide gradients: \[ f(x)=\begin{cases} 0 & \text{if } x < 0 \\ 1 & \text{if } x > 0 \end{cases} \] \[ \sigma(x)=\frac{1}{1 + e^{-x}} \] Through these improvements, AERL significantly improves the performance of ERL systems, especially with the help of GPU acceleration and back - propagation fine - tuning, making the evolutionary algorithm more efficient and practical.

GPU-Accelerated Rule Evaluation and Evolution

Generalize Robot Learning from Demonstration to Variant Scenarios with Evolutionary Policy Gradient

Evolution-Guided Policy Gradient in Reinforcement Learning

GPU-accelerated Evolutionary Multiobjective Optimization Using Tensorized RVEA

GEVO-ML: Optimizing Machine Learning Code with Evolutionary Computation

EVOTER: Evolution of Transparent Explainable Rule-sets

Evolutionary Retrofitting

Adaptive Evolutionary Reinforcement Learning with Policy Direction

EvoX: A Distributed GPU-accelerated Framework for Scalable Evolutionary Computation

Towards Large-Scale Evolutionary Multi-Tasking: A GPU-Based Paradigm

LLM Guided Evolution -- The Automation of Models Advancing Models

GEAR: A GPU-Centric Experience Replay System for Large Reinforcement Learning Models

Neuroevolution of Recurrent Architectures on Control Tasks

Survival of the Fittest: Evolutionary Adaptation of Policies for Environmental Shifts

Human-Readable Programs as Actors of Reinforcement Learning Agents Using Critic-Moderated Evolution

Evolutionary Reinforcement Learning via Cooperative Coevolution

HyperBlocker: Accelerating Rule-based Blocking in Entity Resolution using GPUs

evosax: JAX-based Evolution Strategies

Using GPU to Shorten the Match Time of Rule Reasoning Based on Rete Algorithm

RulE: Neural-Symbolic Knowledge Graph Reasoning with Rule Embedding

AutoEG: Automated Experience Grafting for Off-Policy Deep Reinforcement Learning