NIPS Conference workshop chair Dmitry Storcheus shares his view on career in machine learning

NIPS Conference workshop chair Dmitry Storcheus shares his view on career in machine learning
This post was published on the now-closed HuffPost Contributor platform. Contributors control their own work and posted freely to our site. If you need to flag this entry as abusive, send us an email.

Dmitry Storcheus is an Engineer at Google Research NY, where he is doing scientific work on novel machine learning algorithms. Dmitry has a Masters of Science in Mathematics from the Courant Institute (New York University), and despite his very young age, he is already an internationally recognized scientist in his field of expertise. He has published in a top peer-reviewed machine learning journal, JMLR, and has spoken at the international conference NIPS. Dmitry garnered peer recognition for his foundational research contribution, published in his paper "Foundations of Coupled Nonlinear Dimensionality Reduction", which has been cited by scientists and engineers. He is a full member of the reputable international academic associations Sigma Xi, the New York Academy of Sciences, and the American Mathematical Society. This year, Dmitry is also a primary chair of the NIPS workshop "Feature Extraction: Modern Questions and Challenges".

We often hear that "Data Scientist" is the new dream job in America; as once stated by Forbes, data science is "the sexiest job of the 21st Century". Data scientists use complex machine learning algorithms and distributed computations to learn and predict patterns in data. As shown by the O'Reilly "2015 Data Science Salary Survey", the average salary of data scientists in the USA has rapidly grown above $100,000/year, thus attracting numerous bright young scientists to start a career in this field. Also, according to the O'Reilly survey, the geographical clusters of data science in the USA are Silicon Valley and the New York area. The number of data scientists in the USA has "more than doubled over the last 4 years", as noted by an RJMetrics survey.

As data science and machine learning are becoming an increasingly popular career direction for young, talented mathematicians and computer scientists in the United States, more of them are working hard to publish in peer-reviewed journals and speak at top-level machine learning conferences in order to build their brand and attract attention from leading scientific labs, such as Google Research, Facebook Research, and Microsoft Research. The leading scientific international conferences right now are the Neural Information Processing Systems (NIPS) and the Annual Machine Learning Symposium. The leading engineering conference is DataEngConf NYC. One of the leading journals is the Journal of Machine Learning Research (JMLR). Having the opportunity to speak at those conferences and to publish in JMLR is extremely rare, but honorable and prestigious for young scientists; it demonstrates that their research is recognized by academic peers and opens career paths for them.

I had the pleasure to meet a young and genius scientist, Dmitry Storcheus, who succeeded in this difficult challenge and spoke many times at top conferences and published in JMLR. He earned recognition from the international machine learning community for his research contributions, and now he is working at Google Research. Remarkably, this year he is a primary chair of the NIPS workshop "Feature Extraction: Modern Questions and Challenges". I learned about Dmitry Storcheus' exceptional story of success and recognition, and his views on a research career in machine learning and data science. Dmitry's story will inspire many of our readers who are interested in machine learning.

Figure 1. The geographical distribution of data scientists in the USA. Source: O'Rilley.

Figure 2. The cumulative number of data scientists over time. Source: RJMetrics.

I met Dmitry Storcheus at a recent data science conference, where he told us more about his research contributions and shared his success story. The first thing that shocked me as we started talking was that Dmitry invited me to a conference workshop that he is chairing this year! It is remarkable that he is only 25 years old, but already serves as the primary chair of the NIPS Workshop "Feature Extraction: Modern Questions and Challenges," that is being held on December 11 in Montreal, Canada (check it out at NIPS, also known as the Neural Information Processing Systems Conference, is the largest and the most cited machine learning conference in the world. The workshop chaired by Dmitry is a part of NIPS, dedicated to the particular topic of Feature Extraction. I was curious to ask: How did it happen that Dmitry, being a very young scientist, had gained such a respected position as a primary workshop chair, while the chairs of other workshops are renowned and experienced professors? I took a closer look at the past NIPS conferences and learned that Dmitry Storcheus got his recognition from the NIPS international machine learning community for his paper "Theoretical Foundations of Learning Kernels in Supervised Kernel PCA," which he presented at NIPS in 2014. He presented his new Supervised Kernel Principal Component Analysis (SKPCA) algorithm, which had great impact and caused widespread commentary. The NIPS organizers were impressed with Dmitry's research and high professional expertise in machine learning, so they invited him to be a reviewer for NIPS. After Dmitry had completed the peer review of the papers submitted to NIPS by machine learning scientists from different countries, he was extremely excited to learn that his reviews were valuable for NIPS and critical for the acceptance and rejection decisions. This was especially true because many of the scholarly works that he reviewed were written by the well-known machine learning professors of much more senior level than Dmitry himself. This proved that Dmitry was experienced in judging peer work, which is a necessary skill and expertise required to chair a workshop. Thus, the NIPS organizers approved Dmitry Storcheus to be a principal chair of the Feature Extraction workshop! In December, he is planning to organize and structure this entire event himself; I will definitely attend.

Dmitry started his machine learning research early on in 2011, and so far, he has given talks at more than 10 international conferences. After talking to Dmitry, I was able to get useful advice for the young audience that is going to college or grad school and wants to pursue the machine learning route and participate in the top conferences and research venues in this field. Most importantly, they need to start working on applied research early in their career. For that, selecting a good advisor is vital: they have to look at their potential advisor's publications, venues, and topics, and the research areas explored in general by each advisor. Dmitry Storcheus has been extremely fortunate to have had an excellent advisor, Professor Mehryar Mohri, at the Courant Institute, who could guide his research and introduce him to the best venues to present his work. Professor Mohri is one of the most cited researchers in machine learning; he is famous for his work in speech recognition that is currently used in every mobile device (have you heard of Google Translate?).

Figure 3. Base salary in data science. Source: O'Rilley.

Let's say that you followed Dmitry's advice and now you are working on an interesting applied research topic with your professor - why is it important to get out there and present your research at peer-reviewed conferences? That is needed for building your brand and recognition. A top conference like NIPS or the Annual Machine Learning Symposium is dominated by experienced researchers who have already earned recognition and proved that their research is valuable for society. If a young scientist is invited to these venues and his/her talk is well-received there, it means that his/her scientific contribution is recognized and valued by the global research society. Such recognition is not only a significant honor, but it also pays off further in your career. Dmitry Storcheus had this experience at the 9th Annual Machine Learning Symposium in New York, where the audience consisted of top scientists and professionals from all over the world. He was stressed out, but his talk was well-received and recognized with an Honorable Mention by the Scientific Committee. As a result, after his talk at the Symposium, he received the Best Spotlight Talk Award. This Best Talk award is rare for a young scientist to receive, and is given out mostly to reputed researchers. This proves that Dmitry's original research contribution was recognized by senior scientists in the international machine learning community and by the members of the New York Academy of Sciences. He was happy to see that his efforts had paid off: after receiving the Best Talk Award, he was honored with an invitation to serve as a supporting member for the New York Academy of Sciences.

It is even more critical for a machine learning scientist to publish in JMLR than to get accepted into a top conference. This journal is rigorously peer-reviewed, so that getting your submission accepted is a great challenge. Dmitry Storcheus has two publications in JMLR, scheduled to be printed in a few weeks. He gave us some advice on how to get into such a journal and what it takes for a submission to pass peer review. His advice on this matter is not just mere words; rather, it comes from his years of experience in judging and reviewing the work of his peers. For two years, Dmitry was a reviewer for the NIPS conference. His reviews have been a turning point in making acceptance and rejection decisions. He is much honored to be a reviewer for such a famous international conference as NIPS, since some of the papers that he reviewed were authored by renowned professors and scientists in machine learning. The main conclusion from Dmitry's reviewer experience is that every claim in a submitted paper should be supported by proofs and numbers; the authors must clearly show that their contribution really brings something new to the table. Reviewers expect papers to have large impact: if it is applied work, then algorithms must be scalable, and if it is theoretical work, theorems must be enlightening. Last but not least, Dmitry encourages authors to answer the question "What is your direct contribution in this paper?" briefly in 1-2 sentences. You would be surprised at how many works are rejected because they fail to explain their contribution clearly.

Scientists work hard to contribute to machine learning and earn recognition at top conferences and journals. The goal of these efforts for most people is to become part of a great research lab. Dmitry's active participation in conferences, publications, and awards proves that he has earned recognition in science and has gained him attention by the Google Research lab, where he is now working to apply his research to Google products that most Americans use. As you see, machine learning in America is full of incredible opportunities for young genius scientists like Dmitry. I wish him the best of luck and am excited for him to continue conducting more novel research in the US.

Before You Go

Popular in the Community