USC Professor Dr. Bhaskar Krishnamachari

USC Professor Dr. Bhaskar Krishnamachari : Education, the Internet of Things, and Contact Tracing

USC Professor Dr. Bhaskar Krishnamachari

1. Can you provide some background about your upbringing?

I had a wonderful childhood, growing up in New Delhi, India till I was 15. I spent a lot of it playing cricket, soccer, tennis, reading comic books and pulp fiction, hanging out with friends, watching birds. My parents encouraged me to learn at my own pace, on my own terms. It was not till high school that I got into academic subjects and grew to really love subjects like mathematics, history, science.

When I was 15, I moved to the US, accompanying my father who was an Indian civil servant posted to work at the United Nations headquarters in New York City. I did a year of high school in NYC before starting my undergrad in Engineering at The Cooper Union. I did my graduate studies (MS and PhD) at Cornell till 2002, and since then have been a faculty member at the University of Southern California.

2. What advice would you give to yourself after you graduated from Cooper Union?

I don’t know that I would do anything different than I did – go on to pursue a Ph.D. It gave me a chance to become a self-confident and self-sufficient learner and set me on the path of discovery and learning through research and sharing that learning through teaching and publications, which I have done since then.

3. Why did you choose to pursue research in the Internet of Things and Blockchain Technologies?

I came to IoT through my interest and research going back nearly two decades, to the early 2000’s, on wireless sensor networks. The promise of widely deployed sensors to “reveal the previously unobservable” and improve our understanding of the world was appealing to me.

Blockchain is a somewhat more recent area of interest, going back about 4 years, but I was drawn to it by the recognition that it offers some fundamental new building blocks in terms of greater transparency and trust. It was also an opportunity to build on and grow my understanding of networks, distributed systems, network economics, and applied cryptography.

Something common to both fields is that they are inherently multidisciplinary both in combining many technologies, and in their wide applicability to many domains.

Accenture’s Chief Data Scientist on Deep Reinforcement Learning

4. Which paper that you have written so far have you been most proud of, or believe is most significant? Why?

This is a bit like asking a parent to pick their favorite child 😉 and I’m not sure I could pick just one from a few hundred. But the following three, all written with brilliant Ph.D. students at USC, would rank right up there:

* In 2005, with my PhD student Kiran Yedavalli, we developed sequence-based localization – a fundamentally new kind of wireless signal based localization algorithm that uses rankings of radio signals from (or to) different known positions to determine the location of a wireless device at an unknown location. This can be used as a technique for indoor localization. http://ceng.usc.edu/~bkrishna/research/papers/ecolocationIPSN05.pdf

* In 2010, with my PhD student Scott Moeller, we wrote a paper titled “routing without routes” that showed how you can route data in a mesh network of wireless IoT devices without pre-establishing static routes, and instead moving data dynamically and in a completely distributed manner towards directions that are “less congested”, rather like the way in a kitchen sink water molecules head towards the drain without knowing any “route” to follow. http://ceng.usc.edu/~bkrishna/research/papers/RoutingWithoutRoutes_IPSN10.pdf

* In 2012, with my PhD student Yi Gai, we wrote a paper showing how complex network optimization problems with uncertainty in the underlying variables that can arise in many domains, from computer networks to industrial operations research, can be solved efficiently using online learning algorithms. http://anrg.usc.edu/www/papers/TON-Jan2012.pdf

5. What has been your favorite project as Director of the ANRG, and why? What can you tell us about the future of 5G, and how does American companies’ 5G capabilities compare with those of other countries?

The Autonomous Networks Research Group is, at the end of the day, primarily a collection of brilliant students and researchers I have the privilege to work with. Every project we work on, every student I work with, is unique and thrilling in its own way.

5G is largely a marketing term used by the Cellular industry. But there are a number of interesting research directions underlying that technology, from edge computing to handling much larger numbers of IoT devices that will enable fundamentally new kinds of applications.

I believe a real strength of the USA is in its strong academic research base, consisting of dozens of world-class research universities.

So long as these US universities continue to draw the best and brightest minds from around the world, and our country remains welcoming and inviting to them to stay and live here, it will continue to be a hub for innovation and US companies will continue to stay competitive in the emerging new technologies for 5G and beyond.

6. What’s your favorite aspect of being a Professor of Electrical and Computer Engineering, and Computer Science?

I am so grateful for being in a field with tons of new ideas and interesting tools and techniques that come out on a daily basis, so there’s always something engaging and exciting to learn about and work on. For this reason, it also attracts the best and brightest minds in terms of enthusiastic students and scholars that I get to work with.

7. You mention in your blog post “University Inc.” that you’ve felt the need for more vocational training so that your students are prepared for their careers. How, as a professor, do you find the balance between vocational training and a more liberal arts approach? Is this balance specifically tilted in the engineering field?

I think it is a question of balance. Teachers tend to be influenced by their own personal academic inclinations and want to focus on those in their classes, but we need to bear in mind that many of our students are a) headed to industry and b) may not necessarily share the same background in or love for theoretical topics that some faculty have.

As a result, classes that are too theoretical can seem disconnected and disjointed for students and even if they do well on exams in that class will forget what they learned very quickly.

There has to be a concrete connection with industry and relevant applications and methodologies so that they can connect their learning to real-world problems. And they have to pick up practical skills with respect to software development, hardware design and integration, oral and written communications and teamwork.

However, these skills should not come at the cost of building the necessary technical foundations, and recognizing that innovations require broader and deeper understanding of multiple disciplines, and not merely vocational skills.

Above all, students should leave a University confident in and excited about their ability to continue learning all through their lives.

8. I found the breakdown of the seven perspectives of math education to be quite interesting as a student who loves math. With which perspectives do you identify most strongly in your own learning, and in your teaching of CS and electrical engineering?

In my research and thinking, I have found the “modeling perspective” to be the most appealing and useful to me personally, and I try to introduce it in every class I teach. It emphasizes that mathematics is a tool to understand and analyze all kinds of physical and social phenomena. It also shows that mathematics is a creative discipline, mathematical modeling is really an art.

9. What’s your favorite part of your job as a whole at USC?

I really love teaching and mentoring students. No two individuals are the same in their background and inclinations and passions, but they are almost always enthusiastic and eager to grow and learn. I learn so much from each of them in turn.

10. Do most widely-circulated Coronavirus models use the SIR epidemic model that you explained in your article “Simulating Covid-19 with 6 lines of code — the SIR epidemic model”? How accurate or reliable are these models?

They typically use just a slightly enhanced version of this called SEIR, which has an additional “Exposed” state – where an individual has been exposed already to the virus but is not infectious yet (accounting for an incubation period).

They can be useful for planning and projection purposes, but what we are finding is that there are many unknowns that are hard to estimate in order to model accurately — for example, the rates of infection not only differ from community to community as a function of the underlying movement and encounter patterns but also as a function of government interventions such as “stay at home” and individual willingness to wear masks and wash hands which may be different from culture to culture.

Improving their accuracy requires fine-grained real time information from various sources, which can be hard to come by.

11. What technologies in electrical engineering and computer science have been and can be developed to combat the spread of coronavirus, especially when the country begins to reopen?

There are many directions in which ECE/CS technologies are playing a role.

To name a few, Bayesian models of disease spread and impact, leveraging real-time IoT data sources, digital contact tracing, ML tools to combat misinformation on social media, federated search and query systems for big data analytics, immunity credentials on the blockchain, AI tools for extracting useful knowledge from 1000’s of academic papers.

12. What steps are you taking to pursue either or both of the protocols you proposed in “Mobile app protocols for privacy-sensitive epidemic contact tracing”? How effective do you think these protocols would be, and what issues might they run into?

I made a conscious decision to publish that article as a blog post in the public domain to try and bring it to as wide a public attention, as quickly as possible.

Working through colleagues at the Viterbi school of engineering, we also reached out to engineers in Industry at places like Google and Apple. About a month later, in April, Apple and Google announced that they were collaborating to develop a Bluetooth API to enable precisely such privacy-sensitive mobile tracing.

Though I don’t know any details of their internal process, I have been told that their work was informed by that article, and what I’ve learned of their approach indicates it’s a lot like protocol 2 in that article (utilizing random beacons). Since then I’ve learned of several other related efforts underway around the world and we have compiled and published a listing of more than 40 efforts related to digital contact tracing:

https://docs.google.com/document/d/1bgsNjPcvoz8fpOutEYiqCJpu2PWYHnGqPI3pcK9mUio/edit?ts=5e89a2b0.

I have been in conversation with a couple of those groups, including https://coronavirus-outbreak-control.github.io/web/ and see if there are ways we could collaborate or lend a helping hand.

I believe that in many countries a centralized approach where contact tracing is done in a privacy-invasive fashion will not work, and this kind of a privacy-sensitive approach is a must not merely because it is the “right” thing to do, but because it will increase adoption.

However, figuring out the protocol is the easy part, it needs a lot of other elements to succeed, such as adoption by local public health agencies, mechanisms for such agencies to publish digital certificates of COVID19 infection (necessary to prevent false alarms).

We will need agile leadership and cooperation between governments and private companies over a very short interval for these schemes to have an effect, and that will not be easy.

AI Investing with New York’s Leading AI Firm – Rebellion Research

USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor USC Professor r