Hey Guys,
I am restarting my old blog after more than a decade this new year (Dec 31th).
I know some of my readers have missed me. I used to write a lot on Data Science, Math, AI & lot many things. I wrote a few books in between, gave few guest lectures, wrote few articles. Otherwise, I have been really really busy with a lot of work.
First and foremost, as a motivational piece - Here is my video to Field prize Math video contest - https://www.youtube.com/watch?v=z_pjsJisdHQ&t=90s (Ricci Flow was used as an important tool in solving Poincare Conjecture (unsolved topology problem for 100 years with $1 million prize in 2010). I came up with the first application of Ricci Flow on network data analysis in 2009 and discussed it with Aaron Naber in 2013 before ICM 2014 including Nash Entropy and Perelman Entropy. The geometric flows, Ricci flow (as Hamilton put it - curvature guided diffusion process) have the unique potential to help us how networks & even information flow in a given network evolves over time. This video depicts how Ricci Flow can help to view networks, sub-networks as topological objects & how communities can be detected in the given network using Ricci flow & surgery.)
Also, Please feel free to listen to Marques Brownlee's Vlog on talking about Tech and AI with Google CEO Sundar Pichai (I like to listen to this all the time in the gym & on the walk in free time) - https://www.youtube.com/watch?v=n2RNcPRtAiY
- Wait for 11:30 when Sundar speaks about One laptop Per Child Project by Nicholas Negroponte (MIT).
A couple of years back I did this fun project on the weekends with the last 50 years of papers of all the field medalists & the rest of the mathematicians (a lot of unstructured text data) to predict the field medal for the 2022 event. Again, this was a graph mining & NLP (natural language processing & text mining) exercise to motivate upcoming data scientists to try something new, figure out new problems to solve, probably unleash the power of predictive analytics & narrow AI as well.
https://twitter.com/Prasad_Kothari/status/1447243507710799872
And yes One of the predicted winners - Bhargav Bhatt has already won
1. New Horizon breakthrough prize (Sept 2021 - https://breakthroughprize.org/News/60 ) &
2. Clay Mathematics Award (Dec 2021- https://www.claymath.org/events/news/2021-clay-research-award )
3. While Maryna Viazovska was covered in Quanta Magazine - https://www.quantamagazine.org/a-mathematicians-guided-tour-through-high-dimensions-20210913/
(Might be a bit of luck as well that my natural language processing & ensemble model on Python, Neo4j & my definition of breakthrough went well after a few changes)
This required a lot of
1. Data cleaning
2. Defining the breakthrough of every winner
3. Developing co-author network data
4. look-a-like profiling based on topics, citations, embeddings, and lot of other things
5. Developing topic models through natural language processing, the topology of networks, superimposing topology of networks with topics & changing the definition of a breakthrough with respect to the papers of the winners
6. Deep Divergence Learning & Ricci Flow for community detection for clustering right topics with unsupervised learning
7. Reinforcement learning & ensemble modeling
And A Secret recipe of this predictive analytics project which came through the concept of the topology of social networks & KBGAN: Adversarial Learning for Knowledge Graph Embeddings - https://arxiv.org/abs/1711.04071
8. Ant Colony Optimization to understand amongst research coauthors have shortest path with innovation breakthrough yield & to win field..
By the way, on a different topic, please do check out GeomStats tutorial on Information Geometry by Nina Miolane & Alice Le Brigant - https://geomstats.github.io/notebooks/06_information_geometry.html
A lot of interesting musings to come this time whenever I get time!
No comments:
Post a Comment