Datacast
Episode 45: Teaching Artificial Intelligence with Amita Kapoor
Episode Summary
Amita Kapoor is an Associate Professor in a college at the University of Delhi. She has 20+ years of teaching experience. She is the co-author of various best-selling books in the field of Artificial Intelligence and Deep Learning. A DAAD fellow, she has won many accolades, with the most recent Intel AI Spotlight award 2019 in Europe. As an active researcher, she has more than 50 publications in international journals and conferences. She is extremely passionate about using AI for the betterment of society and humanity in general. Check out her latest book "Deep Learning with TensorFlow 2 and Keras" at this link: https://www.amazon.com/Deep-Learning-TensorFlow-Keras-Regression/dp/1838823417
Episode Notes
Show Notes
- (2:02) Amita described her educational background, studying Electronics at universities back in the 90s. She also professed her love for Asimov’s writings.
- (3:33) Amita talked about her reason to pursue a path of an academic professor.
- (5:13) Amita discussed her Ph.D. titled "Modeling, Design, and Applications of Optical Amplifiers and Long Period Gatings” at the University of Dehli and Karlsruhe Institute of Technology
- (8:38) Amita shared her opinions on how the education of neural networks has evolved in the last 20 years of her teaching career - including the programming language shift from using Fortran and C++ to Python and the importance of learning computer networking and operating systems.
- (14:29) Amita discussed her research that combines the concepts of social network analysis and neural networks to model user behavior in society (read the full paper here).
- (17:57) Amita talked about the process of writing the TensorFlow 1.x Deep Learning Cookbook with Antonio Gulli.
- (21:08) Amita went over TensorFlow Machine Learning Projects, co-authored alongside Ankit Jain and Armando Fandango.
- (23:19) Amita dived into Hands-On Artificial Intelligence for IoT - which discusses different AI techniques to build smart IoT systems, covering practical case studies in personal & home devices, industrial applications, and smart cities.
- (27:50) Amita explained the improvements in TensorFlow 2.0 from its previous version, referring to her book Deep Learning with TensorFlow 2 and Keras - 2nd Edition in collaboration with Antonio Gulli and Sujit Pal.
- (31:33) Amita went over her experience participating in the NASA Centennial Space Robotics Challenge in 2017, in which her team finished in the top 20 out of more than 100 teams worldwide.
- (34:49) Amita reflected on her volunteering experience with a group of friends to build an Acute Myeloid Leukemia detection system that won an award for the Intel showcase in 2019.
- (38:21) Amita unpacked her blog post looking at COVID19 from a data science perspective.
- (40:26) Amita described her mentoring work at Neuromatch Academy, a non-profit online course in computational neuroscience.
- (44:19) Amita shared her opinion on the benefits of an online classroom versus an in-person classroom.
- (51:19) Amita expressed her thoughts on the tech and data community in New Dehli.
- (52:24) Amita shared her hobby of writing science fiction stories.
- (53:41) Closing segment.
Her Contact Info
Her Recommended Resources
Episode Transcription
Key Takeaways
Below are highlights from my conversation with Dr. Kapoor:
ON BEING AN ACADEMIC PROFESSOR
- Back when I was studying in school, I had two passions: reading and teaching. I used to teach my classmates on subjects that they did not understand in exchange for books. Reading and teaching brought me a lot of questions. That’s the whole idea of research — finding answers to your questions.
ON THE EVOLUTION OF AI EDUCATION
- When I started in 1996, the students had a lack of interest in neural networks. Around 2012 when the hype started, everyone wanted to study it. Concerning neural networks' educational aspect, people can learn to write a program, but understanding how the model works still requires an in-depth understanding of the subject.
- Students used to code up multi-layer perceptrons and implement the back-propagation algorithm by hand in Fortran and C++ back in my early teaching days. We were also limited by computational power — simple programs used to take 3–4 days to train.
- Nowadays, with Colab and Kaggle notebooks, the same code can run in 30 minutes. Furthermore, Python is a comparatively easier language for students to get started on, thanks to its interpretability.
ON WRITING TECHNICAL BOOKS
- Presenting the materials to a technical audience in the right manner is very important. Furthermore, I need to make sure that the content is 100% correct, given the deep expertise of my audience.
ON TENSORFLOW 2.0
- TensorFlow 2.0 has eager execution. In the 1.0 version, you have to define the computation graph first and then perform the execution. That’s not intuitive enough to debug error sources. With eager execution, you can build the model without defining the execution.
- TensorFlow 2.0 also has a nice integration with Keras. You can use the model API from Keras and configure your TensorFlow model to have a similar form.
- TensorFlow 2.0 has a library called Strategy for distributed training, where distributed execution takes place by itself without manual configuration.
ON MOVING TO VIRTUAL TEACHING
- Students can get assistance whenever they want in virtual classrooms, as the professors are not bounded by the time.
- The discussion forum in virtual classrooms is an excellent source of knowledge. The teacher is not an expert in everything, so a discussion forum encourages students to share their knowledge and support their peers.