(c) depositphotos.com

5th International Conference on Machine Learning for Networking (MLN'2022)


Paris, France, November 28-30, 2022

Keynote: Soumya Banerjee, Research & Innovation Trasna Solutions Ltd. (Europe)


Title: Interpretable & Explainable Machine Learning (IML/XAI) in the Industrial IoT domain: Are Bayesian Optimization and IML/XAI far from each other?


Abstract

The trend of the IoT enables the generation of massive amounts of data to support inferences and decisions after precise analysis. Most unlabeled data through any machine learning or AI algorithm can extract meaningful insights and provide insightful decisions that will affect human life, society, and the environment. Applications are envisioned from healthcare, reconnaissance, recommendations, and predictive maintenance to the autonomous IoT paradigms. Given these broader ranges of machine learning and AI application areas, it becomes crucial to understand how and why decisions are made. However, as machine learning models become more powerful, they often become more complex and also less transparent. These powerful models often become "black boxes" and suffer from opacity. In other words, they exclude the internal logic of their users/interfaces. Therefore, recently, the concept of Explainable Artificial Intelligence (XAI) and Interpretable Machine Learning (IML) has put more effort to build explainable and interpretable IML/XAI algorithms. Explainable/interpretable AI systems help shed light on the core of black box models that reveal unseen/hidden information such as feature importance and feature correlations. However, only IML/XAI is not independent enough due to a massive amount of IoT data traces. Therefore, following black-box optimization problems such as material design (smart material design), power plant tuning, data center cooling, and hyper-parameter tuning, require optimization algorithms to place the relevant data into ordered dimensions. This Bayesian optimization processing could strengthen IML/XAI processing. This talk will describe the development and deployment of IML/XAI algorithms and their optimization in the field of Industrial IoT while showing a niche experience of these algorithms and their development.

Biography


Dr. Soumya Banerjee (SM-IEEE) is Senior Vice President Innovation, TrasnaSolutions Ltd. (Europe) (https://www.trasna.io/) and Senior industrial Research Fellow ( University College Cork, Ireland, www.ucc.ie He was also an adjunct invited Research professor at Conservatoire National des Arts et Métiers (CNAM), Laboratoire CEDRIC and at present as Associated Researcher at and INRIA– EVA Paris, France (https://team.inria.fr/eva/team/), the French National Institute for computer science ) Paris since November 2018.
Prior to that he was senior Associate Professor, Computer Sc.&Engg., Birla Institute of Technology Mesra, India, visiting research professor at CNRS –INSA de Lyon, Lyon, France (2016), Invited Research professor at TU-Ostrava, Cz Republic respectively (2015). He also spends several years with MSR Seattle, USA, in Cognizant Technology Solution, ICICI InfoTech both in India, south East Asia and Europe.
Dr. Banerjee completed his Bachelors in Engg. (at present VNIT Nagpur) in Computer Sc. (Hons.) and Ph.D in Computer Science and Engg. from Birla Institute of Technology, Mesra, India on Stigmergic optimization with Hybrid Intelligence in 2008-2009. He has more than 130 international journal publications including 34 book chapters and 56 International top level conference proceedings published from Elsevier Science, IEEE Transactions, ACM, Springer– Verlag Germany, CRC Press, and Idea Publication USA to his credit 2 covering machine learning, security measures, prediction and data analytics, bio inspired intelligence, soft computing and optimization, hybrid intelligence, social networking applications and social media, Wiki analysis, machine learning with complex system and evolutionary computing. He also guided more than 14 Ph.D scholars in India and abroad.