Seeking to propel NYU to the forefront of the rapidly developing field of data science, the Initiative in Data Science and Statistics will build on the university’s strength in many fields of knowledge at schools across its academic spectrum, including the Leonard N. Stern School of Business; the Polytechnic Institute of NYU; the Center for the Promotion of Research Involving Innovative Statistical Methodology (PRIISM) at the Steinhardt School of Culture, Education, and Human Development; the Center for Health Informatics and Bioinformatics at NYU Langone Medical Center; the College of Arts and Sciences; and the newly created Center for Urban Science and Progress (CUSP).

The initiative includes two separate but complementary components: education (Master of Science in Data Science program) and research (Center for Data Science). Both are essential because as massive data sets are constantly being generated, it has become increasingly important to extract knowledge from them, requiring both the teaching and utilization of advanced analytic methods.

The difference between big data and data science

Big data and data science may seem similar, even identical, to some, but there is an important distinction between the two. Gérard Ben Arous emphasizes: “We are not doing big data. This is crucial. The difference between the two is the word science. I am, we are, scientists.”

He states, “Big data is more concerned with the engineering components of data and in answering the following questions: how do you store it, how do you manipulate it, how do you do parallelized computations on it, how do you access it, how do you mine it? That is more of what CUSP will be interested in and we will collaborate. But we will do more science,” he says, “looking at the algorithmic and mathematical aspects of extracting knowledge from data.”

-By ML Ball

The post Initiative In Data Science And Statistics: A Collaborative Effort With Two Distinct Components appeared first on Data Science at NYU.

“In April 2010, there were two one-page articles in the journal Nature that are very representative of the two schools of thought that exist in biology and medicine right now,” he notes. “The first takes the traditional approach where someone understands a domain intimately well, such as a gene or disease. This person thinks hard and long and comes up with a hypothesis, then says, ‘Let’s do a test in the lab.’ This is how we practiced science for quite a few years.”

The second approach, he says, is the result of advances in computational sciences and technology. “We now have the ability to generate tons of data in a guided way from a cell, a tissue, or an organism. So the question is, can we figure out what the data is telling us? How can we exploit this unprecedented capability to answer existing questions, as well as break new ground and advance knowledge?”

In the first approach, Rigoutsos explains, the scientist is limited by his or her imagination. In the second, one is limited by technology, and technology has been making great strides. “This is where it gets interesting,” he says, “because we can now generate information about individual cells that we could not have fathomed ten years ago. It challenges you to think in ways you wouldn’t have done if you followed what the books say.”

Commonly, in situations where data says one thing and the books say another, Rigoutsos and his colleagues are faced with a choice: Which one to believe? “My position has always been, believe the data,” he states firmly. “It doesn’t matter what your personal beliefs are, it doesn’t matter what the books say. If the result is repeatable and the experiment has been done correctly, you have to believe the data. You have to learn to liberate yourself from the constraints that come with formal education. As scientists, we are trained to think differently but when it comes to practicing it, it’s not always easy,” he says.

According to Dr. Rigoutsos, there is an increasing realization by the medical community that the new way of doing science, namely, smart processing of big data, will shape our scientific activities for years to come. This, he says, will likely be a long process but will happen. “You need to train new people while also convincing the practitioners who have been doing it a different way that they have more to gain if they open themselves to this new possibility,” he states. “Biology and information science took about 20 years to converge, so we’re probably looking at 20 years if not more for medicine, with the end result that medicine will be practiced in a radically different way.”

By M.L. Ball

The post In The Field Of Computational Medicine, Isidore Rigoutsos Trusts The Data Rather Than The Books appeared first on Data Science at NYU.

Rising global sea level is one of the most pressing issues currently being studied by the CAOS group – specifically, “why it is changing and how much might it change in the next hundred years or so,” explains Professor Holland. “To try to determine that, we look at two huge ice sheets, one in Greenland and the other in Antarctica. When this ice melts and flows into the ocean, it causes the sea level to rise, but not just in one place. Similar to an ice cube being put into and then melting in a glass of water, when marine ice melts, the sea level rises everywhere, really fast,” he says.

Holland’s group of researchers is studying not only the thickness of the ice but also the movement of water flowing out to the ocean and back in to land again. For millions of years, melted ice has been traveling to the ocean by way of canal-like passageways, or fjords, Holland says. “When warm water from the ocean flows through the fjords to the ice-covered land, this causes the ice to melt and come back out. Our goal is to place a temperature sensor into the ocean so we can see whether or not warm water is coming into the fjords and ice is flowing out, which is the issue affecting global sea level change for this century and the next.”

Surprisingly, even with today’s technology, instrumentation, and data analysis capabilities, Holland says that scientists cannot predict future sea levels because the mathematical and physical constructions to simulate what will happen to the Greenland ice sheet, and far more importantly, the larger Antarctica one, do not exist.

“We have mathematical tools, numerical methods and computers,” Holland explains. “But when you attempt to write your equation down, you realize you don’t know what to write because you haven’t studied the details of what you’re trying to describe. That’s why for the last five years, I’ve done field observation, collecting data that will hopefully bring some insight into what we’re trying to describe.”

Collecting this data will take at least 100 years, Holland predicts. “Generations will be doing this. We’re just trying to make moderate progress and meaningful contributions, so that future researchers will find our data set really useful, and can then go off and develop some theory based on it.”

NYU’s Center for Data Science is well-timed, Holland notes, coinciding with the data his group is collecting. “There are vast amounts of data coming from satellites flying over the ice sheets daily, generating vast data sets over decades,” he says. “The ability to sift through this requires data science. This is becoming so important and sophisticated, and it’s allowing us to learn amazing things,” he adds. “I can see only more of it coming.”

By M.L. Ball

The post By Studying Polar Ice Sheets, David Holland Seeks to Predict Rising Sea Levels appeared first on Data Science at NYU.

Commenting on NYU’s new Initiative in Data Science and Statistics, Cappell states, “That’s a great initiative. Things have continued to develop in a number of areas that will benefit from this, within Courant, NYU, and society at large. Human society is overwhelmed by vast amounts of data that need to be organized, understood, and made into the basis of new knowledge,” he says. “Climate science data, economic data, biomedical data, genetic data and others are now available because of new ways of measuring phenomena. Just think of what’s going to happen when patients’ records get linked up with their genomic data, and the database that will be available in terms of health and treatment outcomes.”

The combination of greater instrumentation and greater computerization, says Cappell, means that “in every direction, we are producing reams of data and they need to be comprehended. I think it’s really good that NYU has an exciting initiative in this direction.”

After almost four decades in the field of mathematics, Professor Cappell describes the current climate as one which is not only developing rapidly, but developing on all fronts. “All of science and many areas of mathematics are developing at extraordinary speed, considerably greater than when I was young,” he says. “Historically, there were certain areas of science that were advancing rapidly, and then there were others where you could go to sleep and wake up a generation later and you wouldn’t have missed much. That’s not true anymore.”

In mathematics, Cappell says, as in science more broadly, the number of new areas and the speed of their development has picked up amazingly, partially fed by the ways in which areas create tools that in turn, get used by other areas of mathematics. “My field of topology, for example, has created tools and methods and basic foundational results that have spun off into other fields,” he explains. “This has enabled rapid growth and development in ways that couldn’t have been anticipated.”

By M.L. Ball

The post Courant Institute’s Sylvain Cappell Sees Science, Math and Data Developing at Extraordinary Speed appeared first on Data Science at NYU.

“There is a huge demand in industry for graduates familiar with techniques to extract knowledge from data,” explains Yann LeCun, Director of the Center for Data Science. “Currently you can’t find those people anywhere because they need a combination of expertise in applied mathematics, statistics and computer science – particularly machine learning. There is no program anywhere in the country that educates people with this combination of expertise. Some students manage to acquire the right set of skills by taking courses from various programs, but it’s not easy. The MSDS is designed to cover the right set of topics for a well-rounded data scientist,” he states.

The realm of applications for this degree is much wider than the business world, according to LeCun. “It includes the physical and natural sciences, medicine, engineering, the social sciences, advertising and marketing, government, urban studies…it goes on and on. You need people with a deep theoretical and practical understanding of the methods and techniques. And that’s what our graduate program will be about. Very few programs in the United States are that encompassing,” he says.

The MSDS is a two-year, 36 credit program with six required courses covering computational and mathematical statistics, advanced machine learning, relevant areas of applied mathematics such as large-scale numerical optimization, and questions related to large-scale systems and big data. Among the required courses is a capstone project in which a real-world data science application is designed and constructed. Student must also take six elective courses chosen from a wide range of disciplines and domains, from computer science and the mathematical sciences to fields of application such as finance, business analytics, genomics/bioinformatics, neural science and the social sciences.

The MSDS program should appeal to students with a wide range of backgrounds. “The ideal applicant,” LeCun says, “has a strong background in the mathematical sciences, typical of undergraduate majors in engineering, physics, mathematics, statistics, computer science, economics and other analytical fields, as well as a good knowledge of programming and basic computer science. Ideally, candidates would have taken calculus, linear algebra, a basic probability or statistics course, and a few advanced courses with a heavy mathematical content such as Calculus II and III, or advanced physics, engineering, or applied mathematics courses. They would also have taken two or three courses that involve programming.”

Hosted by NYU’s Courant Institute of Mathematical Sciences, the Center for Data Science will be composed of a number of core faculty, co-located in the center alongside their students and postdocs, as well as a number of associated and affiliated faculty from all across the university who will participate in the center’s research and teaching activities.

More information is available on the CDS website.

Applications will continue to be reviewed until the program is filled.

-by ML Ball

The post Center for Data Science Taking Applications for Masters of Science in Data Science appeared first on Data Science at NYU.

In fact, the computer cluster is already being used this semester in NYU’s CILVR Lab (Computational Intelligence, Learning, Vision and Robotics) course “Big Data, Large Scale Machine Learning,” co-taught by Yann LeCun, Courant Institute of Mathematical Sciences Silver Professor of Computer Science, Neural Science, and Electrical and Computer Engineering, and John Langford, Doctor of Learning at Microsoft Research. The CILVR Lab is part of the NYU Computer Science Department.

“Yahoo! has a long history of working with big data,” said Ken Schmidt, Head of Academic Relations at Yahoo! Labs. “With the torrent of information and data that’s coming to the Web or being accessed through the Web, people are being deluged with tons of data. What they want is information, personalized for their individual needs. That’s what data science means in part to Yahoo! and why the Initiative at NYU is important to us. These initiatives that are being started at great academic institutions like NYU are fostering more faculty and students to focus on problems that are of interest to Yahoo!,” he said.

In addition to contributing the computer cluster, Yahoo! Labs is sponsoring a seminar series to be held at NYU this spring, entitled the “NYU-Yahoo! Seminar Series in Data Science.” By doing so, Yahoo! Labs hopes to encourage “a greater dialogue in the area of data science as well as provide speakers to stimulate conversation about new research projects in data science,” Schmidt explained. “We also hope to get Ph.D. students interested in working on problems in areas where we see the web going. Maybe later, they will want to join Yahoo! and work on these,” he said. “Yahoo! Labs wants to share, as much as possible, data and problems and our scientists with people in academia, and this seminar series is one example of that.”

Schmidt added that Yahoo! Labs is also very interested in NYU’s university-wide Initiative in Data Science and Statistics. “It is a marvelous initiative that will bring together faculty and experts in the areas of statistics, computer science and applied math, and beyond that, artificial intelligence. Currently there is a groundswell around the major universities to foster this kind of interdisciplinary communication and collaboration to find tools and techniques to extract information and trends from huge data sets,” he said. “It’s important and it’s the kind of thing Yahoo! Labs wants to encourage.”

Written by M.L. Ball

The post Yahoo! Labs A Founding Sponsor of NYU Initiative Data Science And Statistics and Center for Data Science appeared first on Data Science at NYU.

In addition to his Managing Director appointment, Lowrance continues to be a Senior Research Scientist in Computer Science at NYU’s Courant Institute of Mathematical Sciences, and is currently a doctoral student in machine learning, also in the Computer Science Department of NYU. He is a founder of Advanced Valuation Analytics Corporation, where he develops real estate valuation and mortgage default prediction models.

Prior to joining NYU, Lowrance was Chief Technology Officer at Reuters and Capital One Financial. He was also a management consultant at McKinsey and Company, where he focused on financial services companies, and a partner at the Boston Consulting Group, where he was responsible for helping clients leverage information for competitive advantage.

Lowrance received his Bachelor of Arts degree in Mathematics from Vanderbilt University, graduating cum laude as well as earning entry into the Phi Beta Kappa Society. He subsequently earned a Master of Business Administration with high distinction from the Harvard Graduate School of Business Administration, and was designated a Baker Scholar, an honor awarded to the top five percent of the class.

Written by M.L. Ball

The post Roy E. Lowrance Named Managing Director of NYU Center for Data Science appeared first on Data Science at NYU.

“It’s great to have John on board because he has considerable practical expertise in these issues,” said Yann LeCun, Director of the recently-launched NYU Center for Data Science. “He’s making his techniques available for the world to see.” The course is a prototype for one of the Center for Data Science’s required core courses in the Master in Science of Data Science to be launched in the fall of 2013.

Formerly at Yahoo! Research and now at Microsoft’s new research lab in New York City, John Langford has been working on various techniques to make machine learning more reliable, predictable, automatic, effective and parallelizable. “I’m a researcher and I believe if you’re doing good research, it should be the kind of thing you teach,” he said. “The class is a great idea because there are a lot of things that have been developed in machine learning. The primary development center is still substantially in academia but there’s also a considerable part now in industry. The fact that I can teach at NYU is very helpful for students because we have various techniques and tools that aren’t as available otherwise in academia.”

With enormous changes in machine learning happening right now, LeCun and Langford feel the course is coming at just the right time. According to Langford, “Machine learning is becoming an industrial tool and is going to stretch in various ways in order to accommodate what’s needed for industry. What’s really driving it is people have lots of data and they want to figure out how to do something useful with it.”

Both LeCun and Langford agree that demand is extremely high for people with this kind of specialized learning. “In this revolution of data-driven knowledge in science, business and government, we need people who really know the methods and techniques for extracting knowledge from data,” said LeCun. Echoing his statement, Langford added, “You really want people who understand computers and you really want people who understand data. And then you want the two of them together because computers and data together allow you to do things you couldn’t imagine doing otherwise.”

To learn more about the course “Big Data, Large Scale Machine Learning,” click here.

Written by M.L. Ball

The post Yann LeCun and John Langford Co-Teaching New Course: “Large-Scale Machine Learning and Big Data” appeared first on Data Science at NYU.

Taught by faculty from across the university, the Center for Data Science two-year master’s graduate degree program will begin accepting applications in February 2013, with classes commencing this fall. NYU’s Courant Institute of Mathematical Sciences will house the Center for Data Science, and Yann LeCun, CIMS Silver Professor of Computer Science, Neural Science, and Electrical and Computer Engineering, will serve as its inaugural Director.

The rapidly-developing field of data science uses automated methods to analyze massive amounts of data and extract valuable knowledge from them. According to LeCun, “The way we are going to do science in the next decade will greatly expand the way we’ve done it so far. We currently collect huge data sets in the physical sciences, life sciences and social sciences. More and more, we need to know how to derive knowledge from them.”

The establishment of the Center for Data Science is both significant and extremely timely, as individuals with expertise in applied mathematics, statistics, computer science and particularly machine learning are currently in high demand. “There is no program anywhere in the country that educates people with this kind of combination of expertise,” LeCun said. “The combination of the NYU Initiative in Data Science and Statistics, the NYU Center for Urban Science and Progress, the Columbia Institute for Data Science and Engineering, and the new Cornell campus in New York City, combined with the high concentration of data-driven companies in and around New York, will make New York City the center of world for data science,” he added.

Written by M.L. Ball

The post NYU Launches Initiative in Data Science and Statistics and Center for Data Science appeared first on Data Science at NYU.

The Norwegian Academy of Science and Letters, along with King Harald and Queen Sonja of Norway, have presented the Abel Prize to the Courant Institute’s Peter Lax in 2005, Raghu Varadhan in 2007 and Mikhael Gromov in 2009.

“‘It really is phenomenal,’” said Donald E. McClure, executive director of the American Mathematical Society, in a 2009 New York Times article. “‘It has the same distinction as a Nobel Prize, and there’s no other institution in the United States or in the world that has had such a concentration of these awards.’”

As part of “Abel in New York,” the Courant Institute’s one-day conference on February 21, 2013, all three Courant Abel laureates will give lectures. Other speakers will include members of the current Abel committee.

Every year, the name of the Abel Laureate is announced in March. In May, the laureate travels to Olso, Norway for two days of celebrations, including a wreath laying ceremony honoring the exceptional Swedish mathematician Niels Henrik Abel in whose memory the Abel Prize was established, a dinner, an award ceremony, lectures at the University of Oslo, and finally, a banquet at Akershus Castle given by the Norwegian government.

The Abel Prize was first awarded in 2003, making Peter Lax, professor emeritus of the Courant Institute, the third Abel Laureate, due to his groundbreaking contributions to the theory and application of partial differential equations and to the computation of their solutions. Asked to describe the award ceremony in Oslo, Lax recalled, “The King was cruising on the Mediterranean but the Queen presided, and the Crown Prince, Haakon, was a very nice guy. He told me he had been a student for some years at Berkeley. He was very pleased with that. It was quite surreal, and a great honor. I was not expecting it,” he said.

Raghu Varadhan, honored for his work in probability theory, said of the experience, “It was really wonderful. People often compare the Abel to the Nobel Prize but with the Nobel Prize, you’re one of seven or eight people in different disciplines, whereas the Abel Prize is just for mathematics, so the entire celebration is just for you. That made it much more impressive for me.”

“I didn’t see it coming,” recalled Mikhael Gromov, awarded the Abel Prize for his revolutionary contributions to geometry. “It was more relaxed and less formal than I was afraid it would be. Seeing the Akershus Castle was really exciting, plus meeting some really interesting people. One change it has brought is more interviews. I knew Grigori Perelman personally so I’ve had many interviews concerning him since winning the Abel.”

Written by M.L. Ball

The post Three Courant Institute Abel Prize Winners To Speak At NYU Conference appeared first on Data Science at NYU.