Abstract:In this research, we investigate the possibility of applying a search strategy to genetic algorithms to explore the entire genetic tree structure. Several methods aid in performing tree searches; however, simpler algorithms such as breadth-first, depth-first, and iterative techniques are computation-heavy and often result in a long execution time. Adversarial techniques are often the preferred mechanism when performing a probabilistic search, yielding optimal results more quickly. The problem we are trying to tackle in this paper is the optimization of neural networks using genetic algorithms. Genetic algorithms (GA) form a tree of possible states and provide a mechanism for rewards via the fitness function. Monte Carlo Tree Search (MCTS) has proven to be an effective tree search strategy given states and rewards; therefore, we will combine these approaches to optimally search for the best result generated with genetic algorithms.

What problem does this paper attempt to address?

The main objective of this paper is to explore a novel approach to optimize neural network weights, specifically by combining Genetic Algorithm (GA) and Monte Carlo Tree Search (MCTS) to achieve this goal. ### Research Background Genetic Algorithm is an optimization technique based on the principles of natural selection and genetics, widely used in machine learning, research and development, and system optimization. However, when applied to neural network weight optimization, traditional genetic algorithms may encounter issues such as high computational cost and large search space. ### Research Problems The problems the paper attempts to address include: 1. **Improving Optimization Efficiency**: Reducing computational resource consumption without sacrificing the quality of the final solution. 2. **Ensuring Weight Effectiveness**: New weights generated through genetic operations (such as crossover and mutation) may perform poorly initially but could improve after several generations of evolution; conversely, the opposite could also be true. Therefore, determining the optimal solution becomes a significant challenge. 3. **Maintaining Weight Integrity**: Avoiding invalid or inefficient weight combinations caused by random mutations. ### Solution To address the above challenges, the authors propose the MCTS-GA method, which utilizes MCTS strategies to guide the search process of the genetic algorithm. This method not only effectively explores the entire tree structure generated by the genetic algorithm but also balances exploration and exploitation to a certain extent, thereby finding the optimal solution more quickly. ### Dataset and Experimental Design The paper uses a diabetes prediction dataset for experimental validation. This dataset, sourced from the National Institute of Diabetes and Digestive and Kidney Diseases, is used to diagnose whether patients have diabetes. By constructing a feedforward neural network with 4 hidden layers and applying MCTS-GA to optimize its weights, the results show that this method improves classification accuracy compared to traditional neural networks and classical genetic algorithms. ### Conclusion Experimental results indicate that MCTS-GA performs well in optimizing neural network weights. Although the performance improvement is not substantial, the computational time is comparable to the other two techniques. There is still room for further improvement in the future, and more testing and validation on different types of problems are needed.

MCTS guided Genetic Algorithm for optimization of neural network weights

Genetic algorithm for neural network architecture optimization

Optimizing Feed-Forward Neural Networks Using Cascaded Genetic Algorithm

Using Genetic Algorithms to Optimize Artificial Neural Networks

Extending MLP ANN hyper-parameters Optimization by using Genetic Algorithm

Studies on Optimization Algorithms for Some Artificial Neural Networks Based on Genetic Algorithm (GA)

Optimization Study of BP Neural Network Based on Genetic Algorithm

Neural network crossover in genetic algorithms using genetic programming

Compute-Efficient Neural Network Architecture Optimization by a Genetic Algorithm

Genetic optimization algorithms applied toward mission computability models

An Aggressive Genetic Programming Approach for Searching Neural Network Structure Under Computational Constraints

Deep Genetic Network

A Multifactorial Optimization Paradigm for Linkage Tree Genetic Algorithm

Genetic Algorithm: An Approach on Optimization

An optimizing BP neural network algorithm based on genetic algorithm

Simulation-based optimization of a production system topology -- a neural network-assisted genetic algorithm

Using Genetic Algorithm to Optimize BP Neural Network

Dual Monte Carlo Tree Search

Genetically Modified Wolf Optimization with Stochastic Gradient Descent for Optimising Deep Neural Networks

Searching Search Spaces: Meta-evolving a Geometric Encoding for Neural Networks

In-Depth Case Study on Artificial Neural Network Weights Optimization Using Meta-Heuristic and Heuristic Algorithmic Approach