Providing effective solutions to global challenges
Tuesday, Oct. 19, 2021: Canadian University Dubai (CUD) is once again demonstrating its ability to lead international research projects with the recent authoring of “Comparative Analysis of Activation Functions in Neural Networks", an article led by Dr. Firuz Kamalov, Associate Professor, Department of Electrical Engineering, Canadian University Dubai. His team will present their findings at the 28th edition of the Institute of Electrical and Electronics Engineers (IEEE) International Conference on Electronics Circuits and Systems (ICECS) to be held Nov. 28 to Dec. 1, 2021, in Dubai, United Arab Emirates.
Collaborating with Amril Nazir (Department of Information Systems, Zayed University, Abu Dhabi, UAE), Murodbek Safaraliev (Automated Electrical Systems Department, Ural Federal University, Yekaterinburg, Russia), and Aswani Kumar Cherukuri (School of IT & Engineering Vellore Institute of Technology, Vellore, India) they investigated the effects of the different activation functions on the geometry of the loss function of neural networks. In particular, they dive into the relationship between these activations in the hidden and output layers, the geometry of the trained neural network model, and the model performance. By presenting visualizations of the trained neural network models, they aim to help researchers better understand and intuit the effects of activation functions on the models.
ICECS is a premier international conference dedicated to circuits and systems, design, and innovations. Held annually in Region 8 (European Union, Africa, and Middle-east), the event will bring together industry leaders, students, researchers, scientists, academics, and many more under the theme of “Security of Microelectronic Circuits and Systems” to facilitate discussion in various topic areas while expanding their network.
Dr. Firuz’s recent work highlights CUD’s continuous efforts to engage in collaborative cross-border research and provide effective solutions to both regional and global challenges whilst expanding perspectives in learning and application.
Recommended Links
- IEEE ICECS Conference