Vishal Deshwal

Research Student - PhD

Research Program: International Centre for Neuromorphic Systems

Thesis Title

Periodic Activation Functions in DNNs.

Research Project

In Artificial Neural Networks (ANNs), activation functions play an important role in interpreting complex patterns and non-linear relationships. Conventional approaches rely on monotonic activation functions. This research seeks to advance the understanding and application of Periodic Activation Functions (PAFs) in ANNs, potentially unlocking new possibilities for network design and performance optimisation. A key focus will be investigating how using PAFs allows us to create multiple hyperplanes for separating or classifying data points, using just a few layers. This approach can allow for the creation of wider networks rather than deeper ones, leading to numerous advantages, such as stability and faster convergence.

Qualifications

• Master of Data Science, Western Sydney University, Sydney

Publications

1. Dimension Reduction in the Sagittal Plane for Diagnosis of Mild Cognitive Impairment - IACC 2022
2. Applicability of deep neural networks & local binary pattern for classification of MCI-C & MCI-NC - Alzheimer's Association
3. Diploid Genetic Algorithm-Based Pipeline for the Conversion Prediction of Mild Cognitive Impairment
4. Multi-Feature Fusion for Mild Cognitive Impairment Staging: Fuzzy Logic, Micro- Structure Distribution and Correlation Information

Contact