Machine Learning, Optimisation, Deep Learning, Network Science, Large-Scale Learning, handwriting recognition, supply-chains, air transport, Mathematical modelling and solving of problems using Mathematical Programming, Meta-heuristic approaches and Reinforcement Learning
Digitally Optimised Through life Engineering Services (DO-TES)
Partners: Industrial partners are Ansys, BAE Systems, Rolls Royce, Bombardier, and academic partners are Cranfield University and City University of London.
Tools/Languages: Python, C, Gurobi
- Applied and published network science approach to study the impact of nestedness on ripple effect in supply-chain networks.
- Applied and published (submitted) optimisation techniques to model the supplier order assignment problem and solved the same using mathematical programming, genetic algorithms and reinforcement learning approaches. The model was successfully deployed by a manufacturing company and it helped to reduce the procurement cost significantly.
Airline Performance and Disruption Management Across Extended Networks (APEMEN)
Partners: Boeing and airlines, like Swiss Airlines, Emirates Airlines and Aegean Airlines etc.
Tools/Languages: Python, networkx
- Applied and published (submitted) data cleaning and analysis of root-causes and impact of flight delay propagation
- Applied and published (submitted) analysis of bottleneck airports, flights and connections to improve airline operations
Solving Large-scale Machine Learning Problems (PhD Thesis)
Tools/Languages: MATLAB, C/C++
- Published and reviewed the big data challenge in machine learning, recent research directions and different areas to tackle the challenge.
- Proposed and published mini-batch block coordinate optimisation framework to solve big data problems.
- Proposed and published variance reduction techniques, viz., SAAG-I, II, III and IV using first order optimisation methods to solve large-scale problems.
- Proposed and published simple sampling techniques to reduce the overall training time of large-scale learning problems.
- Proposed and published stochastic second order optimisation method, namely, STRON to deal with the large-scale learning problems.
- Developed a C++ based library, LIBS2ML (in progress), for scalable second order machine learning algorithms.
Project: Deep Learning for Indic Scripts Handwriting Recognition (online & offline)
Languages/Tools used: Python, Keras
- Proposed and published CNN based architecture for faster (in a few minutes) online Gurmukhi handwriting recognition.
- Working on Gurmukhi handwriting recognition by converting online data to offline (images) and using image augmentation and transfer learning.
Dr Vinod Kumar (Chauhan) is currently working as a Research Associate in Industrial Machine Learning with Institute for Manufacturing, Department of Engineering at University of Cambridge UK. At Cambridge, he has applied network science approach to model airline networks and identified the bottleneck flights, airports and flight-connections to help airlines’ to improve their operations. He has also applied network science approach to study the impact of nestedness on ripple effect in supply-chain networks. Currently, he is applying optimisation techniques to model and solve the supplier order assignment problem. Moreover, Vinod has a PhD in solving large-scale machine learning problems where he has developed optimisation techniques to tackle the big data challenge in machine learning. He has solved the challenge by proposing stochastic variants of first and second order optimisation methods, and also proposed simple sampling techniques to reduce training time of models.