AI Saturdays Bangalore Chapter — Week 4 Reflections

If you could have walked across the third floor of Spring Board, Kormangala, Bangalore and entered the hall at the end, on the 8th of September 2018, I guarantee you would have been perplexed by the sight. 20 odd tables with almost 60 people sitting around these tables and a huge screen with content being projected on it would have greeted your eyes. But that’s not what would have perplexed you. Rather, you would have been amazed by the people themselves. Some of them were young undergraduates while others were working professionals with couple of years of experience on their belt. You also would have found people whose hair might have turned white but with an enthusiasm as infectious as an 8-year-old kid. When your eyes would have travelled to the front of the screen, you would have seen youngsters (read organizers) belting out some code snippets and jargon. Interesting huh? Welcome to the fourth week of Bangalore Chapter of AI Saturdays!

The present cycle of AI Saturdays (refer here for more details on AI Saturdays) for the Bangalore chapter has been carefully planned and divided into three phases. The first phase covered content from and stretched over 3 weeks. The main aim of the first phase was to build sound fundamentals in deep learning and neural networks in particular. The second phase aims to build on these concepts by covering content from courses and the fourth session (session 4) marks the beginning of the second phase of our cycle.

For this session, we had more than the usual number of newcomers (yay !!) and hence, started off the proceedings with a small introduction on AI and the structure of neural networks. Thereafter, basic Python fundamentals like lists, lambda functions, classes and inheritance were discussed. The code snippets related to these concepts were run on Google Colaboratory which provided the necessary exposure to the newcomers. A couple of interesting videos were also played out, the first of which was by 3Blue1Brown on neural networks. We wanted the audience to visualize and get the intuition of the working of a neural network and it is the perfect resource for that purpose. The second video was by Caroline Chan titled “Anybody Can Dance” and is a fantastic example for a research application of neural networks.

Post lunch, we started on the first lecture of Concepts on gradient descent, convolutions and activation functions were examined culminating in a practical session on image classification. This programming session revolved on classifying images on the basis of whether they were of dogs or cats. Thereafter, it was extended onto apples and oranges for reinforcing the methodology. More importantly, a script to download image data automatically (based on certain specifications) was shared with the participants. This will hopefully motivate more people to play with data (The relevant resources can be accessed here ). For this particular session, we would also like to express our gratitude to Springboards for being a gracious host.

Post-lunch programming session: Image Classification with Neural Networks — part 1 — Video 1

The experience of conducting these meet-ups has been great for me and the other ambassadors of the Bangalore chapter. We are bowled over by the passion of the participants and their consistency in the meet-ups. It has prompted us to strengthen the community and the discussions in different ways. While we will be sharing the details of those later, we will still be covering the videos from the curriculum with in-depth discussions and rigorous code implementations in the upcoming sessions.

1. Sign up here to attend next meetups.

2. The resources related to the meetup sessions can be found on our GitHub repository. Do have a look at the roadmap.

3. Assignments for the sessions conducted till date can also be found here .

4. Previous sessions’ discussions can be found here, here, here and here.

5. Follow AISaturdays Bangalore on Twitter and Slack.

6. For any queries, you can contact me on LinkedIn.

Cheers !

Source: Deep Learning on Medium