Explore Machine Learning (ML) is a Google-sponsored program for university students to get started with Machine Learning. The curriculum offers 3 tracks of Machine Learning content (Beginner, Intermediate, and Advanced) and relies on university student facilitators to train other students on campus.
Under this program, a 3-day workshop was undertaken by IET on Campus, BIT Mesra from 8th August to 10th August 2019 in Lecture Hall-1. Shivam Kumar, bestowed with the facilitation of Explore ML, carried out the practicum gracefully with his unique teaching approach along with Kiran Muthigi.
The first day of the workshop commenced at 5:15 P.M. The entire hall was jam-packed with tech enthusiasts. Unfolding the beginners’ track, the attendees were introduced to the basic concepts of Machine Learning, Deep Learning and Neural Networks along with the various types of Machine Learning- Classification, Clustering, Regression, Sequence prediction and Transfer learning. When the day wrapped up at 7:30 P.M, the facilitator had fed enough knowledge to the students for them to describe the meaning and purpose of Machine Learning.
The next day began at the same time and place as the previous one, with participants showing no less fervour than before. The workshop transitioned from beginner track to intermediate track. Students were taught about hidden layers and techniques of using optimum number and architecture of hidden layers. These layers prevent underfitting and overfitting, regularization, learning rates and weight allocation. The facilitator also explained to the learners the step-by-step methodology to make a simple text classifier with TensorFlow hub.
The final day of the workshop commenced at 9:00 A.M. and continued till 5:00 P.M. with a lunch break in between. Students constructed a basic classification model that could differentiate between a dog and a cat. In the process, they also learnt about training a model and achieving greater test accuracy. Further, they were introduced to the concept of CNN or Convoluted Neural Networks and how one could create models with fewer connections between neurons whilst enhancing its accuracy. The final topic of the session was ‘transfer learning’, which involved making use of a pre-trained model to reduce workload.
The 3-day colloquium culminated with a highly inclusive discussion between the facilitator and the participants on possible applications of AI and Machine Learning along with various project ideas involving Deep Learning, Neural Networks and CNN. The ideas included lie detector, an app for automated answer sheet checking, meme recommender, an anti-bully app for social media and many such beneficial ideas. The forum was a success owing to the efforts of Shivam Kumar and the endless hunger for knowledge in the students.
Reporter – Vaibhav Raj Singh
Editor – Utkarsh Vardhan
Picture Credits – The Photographic Society, BIT Mesra