2019/10/06 by Balakrishnan, Harikrishnan Nellippallil, Aditi Kathpalia, Kathpalia, Aditi +4 · 2 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · #Chaos-based Image/Signal Encryption #Chaotic Dynamics (nlin.CD) #FOS: Computer and information sciences #FOS: Physical sciences #Fractal and DNA sequence analysis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Neural Networks and Applications
paper · pdf · doi:10.48550/arxiv.1910.02423
openalex publication_date 2019/10/06 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Inspired by chaotic firing of neurons in the brain, we propose ChaosNet -- a\nnovel chaos based artificial neural network architecture for classification\ntasks. ChaosNet is built using layers of neurons, each of which is a 1D chaotic\nmap known as the Generalized Luroth Series (GLS) which has been shown in\nearlier works to possess very useful properties for compression, cryptography\nand for computing XOR and other logical operations. In this work, we design a\nnovel learning algorithm on ChaosNet that exploits the topological transitivity\nproperty of the chaotic GLS neurons. The proposed learning algorithm gives\nconsistently good performance accuracy in a number of classification tasks on\nwell known publicly available datasets with very limited training samples. Even\nwith as low as 7 (or fewer) training samples/class (which accounts for less\nthan 0.05% of the total available data), ChaosNet yields performance accuracies\nin the range 73.89 % - 98.33 %. We demonstrate the robustness of ChaosNet to\nadditive parameter noise and also provide an example implementation of a\n2-layer ChaosNet for enhancing classification accuracy. We envisage the\ndevelopment of several other novel learning algorithms on ChaosNet in the near\nfuture.\n