A Novel Neuroplasticity-Based Approach for Enhanced Generalization in Computer Vision

Abstract:

In the rapidly advancing realm of artificial intelligence, bridging biological processes with computational models has become a pivotal strategy for enhancing robustness and adaptability in neural networks. This paper introduces and evaluates a biologically inspired neural network layer—termed BioLogicalNeuron—across multiple image classification benchmarks, including CIFAR-10, MNIST, and Fashion-MNIST. By incorporating calcium-dependent regulation, homeostatic control, and adaptive repair strategies, the architecture consistently outperforms standard deep learning methods.

Key Results:

• CIFAR-10: 90.42% accuracy (3.77% improvement over baseline)
• MNIST: 99.43% accuracy
• Fashion-MNIST: 93.27% accuracy

The BioLogicalNeuron maintains stability metrics above 0.92 through targeted interventions triggered at a repair threshold of 0.8. These findings underscore the transformative potential of biologically-derived computational methodologies in augmenting deep learning architectures for image classification, exemplifying a powerful synthesis of neurobiological insights and computational efficacy.

Authors:

• Md Azizul Hakim (Lead Author, Bangladesh Sweden Polytechnic Institute, Bangladesh)
• Rashedul Arefin Ifty (International Islamic University Chittagong, Bangladesh)
• Khaled Eabne Delowar (International Islamic University Chittagong, Bangladesh)
• Samiul Azam Shuvo (International Islamic University Chittagong, Bangladesh)
• Zarin Tasneem (International Islamic University Chittagong, Bangladesh)
• Mohammad Ifazul Alam (Bangladesh Sweden Polytechnic Institute, Bangladesh)

DOI: 10.1109/QPAIN66474.2025.11172203

Key Contributions:

• Introduced BioLogicalNeuron architecture with calcium-dependent regulation and homeostatic control
• Demonstrated enhanced generalization across multiple visual benchmarks
• Established repair threshold methodology maintaining stability above 0.92
• Advanced biologically-inspired computational methodologies in computer vision

Recommended citation: Hakim, M.A., Ifty, R.A., Delowar, K.E., Shuvo, S.A., Tasneem, Z., & Alam, M.I. (2025). "A Novel Neuroplasticity-Based Approach for Enhanced Generalization in Computer Vision." IEEE Conference Proceedings. DOI: 10.1109/QPAIN66474.2025.11172203
Download Paper