Webinar on Embedded Machine Learning in Robotics

#IEEE #UET

Machine learning has become quite the trend in the fourth industrial revolution and is not fizzling out any time soon. The robotics industry is no stranger to applications of Machine Learning. So an event was conducted to guide the freshers about how it really works. Demonstrated the working, scenarios and the working principles of the applications of Machine Learning and AI. Uon Liaquat, a very highly skilled expert in writing custom ML algorithms for learning robots so that they don’t need to be explicitly trained for each task. He is a student of UMT and is currently working on his new startup having theme Reinforcement learning in Robots. The main agenda of the topic of discussion was how to find a solution for a problem in ML and the steps involved for that purpose and to brief the students how ML is really figured out and what they require to do for that purpose.

Uon Liaquat divided the informative and interactive session into three major topics: Supervised Learning, Unsupervised Learning, and Reinforcement Learning. Explained in detail through instances the phenomena, and working of supervised, unsupervised and reinforcement learning. Urged to make the session interactive for a better learning process. Explained different ways to reach a solution and develop multiple approaches to implement Machine Learning in Robotics.

He demonstrated how Supervised Learning is implemented in the problems where you already have a bunch of data and you know the result you want to implement.

Unlikely, unsupervised learning is used where you have a data set but you don’t know what to implement.The speaker further explained that Reinforcement Learning (RL) is an area of machine learning concerned with how software agents ought to take actions in an environment in order to maximize the notion of cumulative reward. Reinforcement learning is one of three basic machine learning paradigms, alongside supervised learning and unsupervised learning. Research work on Reinforcement learning is still in process.

The speaker also explained how to collect data sets from simple models and how are these utilized in machine learning. Algorithms should be coded using mathematical techniques before developing or importing framework. Explained different approaches in manipulating data, parallel processing and the embedding of algebraic and mathematical techniques in the backend. 

Forbade to use frameworks in implementing the solution and start from scratch. Advised to be goal-oriented in order to reach the core of Machine Learning. The speaker conducted a Q&A session at the end of the session and cleared the queries of attendees.