By Sriram Rajamani, Distinguished Scientist and Managing Director, Microsoft Analysis India at Microsoft
Manik Varma, researcher within the Microsoft Analysis (MSR) India lab was awarded the distinguished Shanthi Swarup Bhatnagar (SSB) prize for 2019. The SSB prize is a coveted prize for multidisciplinary science and engineering in India. He was additionally elected to the Indian Nationwide Academy of Engineering the very subsequent day. These honors add to a protracted listing of recognitions that researchers from Microsoft Analysis (MSR) India have obtained over time, together with:
- SIGCHI societal impression award in 2017: Indrani Medhi-Thies, for work on interfaces for low literate customers
- Shanti Swarup Bhatnagar prize in 2016: Venkat Padmanabhan, for pioneering work on indoor localization, smartphone based mostly sensing, and cellular communication.
- MacArthur grant in 2016: Invoice Thies, for advancing the socioeconomic well-being of low-income communities within the growing world via progressive communication and digital applied sciences that reply to real-world constraints.
- Polya prize in 2014: Nikhil Srivastava (who’s now a college member at UC Berkeley) and his collaborators for resolving a long-standing open drawback referred to as the Kadison-Singer conjecture
- Knuth prize in 2011: Ravi Kannan for with many highly effective new algorithmic methods, which have enhanced our understanding of computational complexity, discrete arithmetic, geometry, and operations analysis.
- Pc Aided Verification award in 2011: Sriram Rajamani (with Tom Ball), for contributions to Software program Mannequin Checking
These, and the opposite honors our researchers have received over time are certainly a supply of nice pleasure for us. On the similar time, we acknowledge that just about none of this could have been potential with out the nice collaborations we’ve with our educational companions throughout the globe. MSR, since inception, has at all times acknowledged the significance of collaborations with educational companions and different researchers, not just for advancing the state-of-the-art, but in addition for his or her incisive suggestions on our analysis high quality and instructions. In truth, MSR India has a Technical Advisory board that options a few of the world’s main teachers to advise us on exactly this- the standard and route of our analysis.
Let me use Varma’s (our most up-to-date honoree) work for instance. He bought his PhD from Oxford College in Pc Science with specialization in pc imaginative and prescient in 2004, did a quick postdoctoral stint at UC Berkeley, and joined our lab quickly thereafter. His preliminary work at MSR India continued to be on pc imaginative and prescient–he labored on detecting objects in photos and received competitions within the space. However he quickly converted to Machine Studying and the majority of his work up to now decade has been in Machine Studying. He has achieved work on native deep kernel studying, which obtained each educational recognition and located sensible utility; a virus classifier based mostly on this work is broadly used.
Nonetheless, his most spectacular and impactful of contribution over the previous decade is in Excessive Classification- a sub-discipline of Machine Studying which Manik helped originate. As an alternative of conventional classification algorithms, which classify objects with a small variety of labels (classes), Varma has been constructing classifiers which classify objects with a big area of labels- assume tens of millions of labels! This formulation with a big area of labels is smart for a number of situations comparable to phrase choice (for advertiser bidding, for instance) in on-line commercials, and tagging Wikipedia articles with tags, the place the area of phrases or tags may be very giant. The preliminary algorithms used an ensemble of timber with a lot of methods to assist scale whereas sustaining accuracy. Newer algorithms are based mostly on embeddings generated by Deep Neural Networks.
He began this space and, along with colleagues at Microsoft and college students at IIT Delhi, wrote the primary papers on this subject. Over the previous decade, we’ve seen papers by Varma and his collaborators on this subject routinely showing in conferences comparable to ICML, KDD, NIPS, SIGIR, WSDM and WWW, with a few of the papers profitable greatest paper awards and receiving appreciable consideration and recognition from the group. Greater than ten workshops have been organized within the final seven years in numerous conferences on this subject, and he has been the keynote speaker in most of those workshops.
He additionally collected benchmarks and created an “Extreme Classification Repository”, which has been considered and downloaded 1000’s of occasions each month and is now used as an ordinary to measure algorithmic progress on this space. As well as, he has been a key contributor in constructing ML algorithms that run on very small gadgets, with very low reminiscence and energy footprint.
Varma is an envoy for Microsoft Analysis and the values and rules our group stands for. A lot of the inspiration for his work comes from the real-world issues confronted in industrial settings. He’s additionally an adjunct school member at IIT Delhi, and far of this work was achieved in collaboration with PhD college students at IIT Delhi. Thus, Varma’s award is a celebration of not solely his work, however the collaborations he has had with academia in addition to colleagues at Microsoft.
We reside in fascinating occasions the place business and academia have a fantastic alternative to work collectively to make scientific progress. Some issues that come up in industrial follow are open ended and want a few years of sustained analysis to make progress in fixing them. Publicity to such issues vastly advantages college students and college in academia and therefore such collaborations are mutually useful.
MSR is dedicated to persevering with collaborations with our educational colleagues, to show and formulate tough issues that come up in industrial settingsand bringing to bear the collective knowledge of the tutorial group and advancing the cutting-edge to resolve such issues.
For extra info on MSR India, employment alternatives and collaborations, go to our website.