Author: zzlalo

A Winding Path from Complex Analysis to Computational Biology

zzlalo Career Information, Opportunity, Skill Transfer ,

by Robert Thurman, Principal Computational Biologist,  

Seattle Genetics,

Bothell, WA

Thurman 300

Computational biologists come in two types: those who were originally trained mathematically or computationally and then gravitated towards biological problems, and those who were formally trained on the biological side but couldn’t stay away from computers. That generalization is slightly dated, because many colleges now offer interdisciplinary degree programs in computational biology and bioinformatics, but those programs tend to be small and it is safe to say most practioners currently in the field started out as one of the two types. Both perspectives are important.

I started with bachelor’s degrees in mathematics and computer science, and then took a break between my Master’s and PhD to work as a programmer for NASA’s Jet Propulsion Laboratory. I felt a call back to mathematics, and after following a traditional academic path from PhD to post-doc to tenure-track university teaching position, I switched gears again and took a programming position in the research division of a statistical software company. It was there I was exposed to machine learning techniques applied to biological problems, setting a course towards my current career in computational biology. There are many such winding paths into the field.

Recently I gave a presentation in Research Forum, which is a weekly opportunity at my company to share current results with the rest of the research community. We make targeted therapies for cancer, and I was hired to establish a devoted computational biology function. The forum was the first opportunity to try to neatly summarize what we do as computational biologists. It was a challenge. On the one hand most people in the audience had some exposure to our work, because we collaborate with every group in research. And some functions are well-established and well-known — we do a lot of genomics, for instance, trying to untangle which genes are regulated under treatment, or which ones might presage resistance to therapy. But the nature of our role is also highly varied. In some ways we are analytical “fixers,” and we are happy to take on any kind of problem related to data analysis. In trying to concisely categorize this type of work for my presentation, the best I could come up with was…”Math.”  It’s maybe a bit far from my PhD in complex analysis, but a definite path can be traced back to those roots. And there is a lot of math in this work, albeit in service to a specific (and valuable) purpose.

It’s truly an exciting time to be working in the field of computational biology, especially as it is applied to finding treatments for devastating, tough-to-treat diseases like cancer. Advances in biological understanding and experimental capabilities on the one side, and computational capacity and algorithmic sophistication on the other, have opened the way to new treatments and new tests to get the best therapies to the right patients. Breakthrough advances like immunotherapy have dramatically changed the prognosis for some patients. Advanced non-small-cell lung cancer (NSCLC), for example, has a terrible prognosis, with a 5 year survival rate near 0% for more advanced cases1. But so-called checkpoint inhibitors like nivolumab and atezolizumab, which target the cell surface proteins PD-1 and PD-L1 and free up the body’s immune system to attack cancer, have in some cases doubled overall survival rates compared to previous standards of care2. This level of improvement is virtually unheard of for new cancer therapies, and it means that the field now cautiously uses the word “cures” in cases it never could before. However, only a subset of patients respond to this type of therapy. So the race is on to 1) find biomarkers, that is, some measurable patient characteristics that can predict who is most likely to respond or not respond; and 2) find other immune checkpoints that are successfully druggable.

Computational biology and bioinformatics have prominent roles to play in both of these endeavors. The search for biomarkers involves sifting through data in which dozens to thousands of variables are collected on patients: from height, weight, age and gender, to the number, length and types of previous treatments, to genomic features like gene expression and mutations measured across potentially hundreds or thousands of genes. All of these patient characteristics are then compared to clinical results to see if any variable, alone or in combination with others, could be related to response. Because it is often the case in these types of problems that there are more variables than patients, modern machine learning techniques, such as regularization and random forests, can be used to overcome the limitations of under-determined systems and identify which variables are most important in predicting response. In my own work I use these techniques as well, to try to understand, for instance, what measurable characteristics of our drugs (which are fairly complicated in their mechanisms of action) contribute most to their potency in an in vitro setting. (This would be a good place to add, as a general recommendation to others as well, that I wish I had taken more statistics!)

Finding new immune checkpoints is a special case of the general problem of finding new drug “targets”. This usually means identifying a host molecule like a protein or gene product that is in some way important for the progression of a disease, and whose function can be altered or co-opted with a drug. Computational biology contributes in important ways to this as well. While a traditional approach to finding new targets might be to follow up on a research article that addresses some specific fundamental biology, modern data mining techniques can be applied to vast public data resources like the The Cancer Genome Atlas (TCGA)3 to scan the entire genome, across all cancers, for genes that are, say, preferentially expressed in cancer compared to normal tissue.

Such an exercise ties directly into one of the pleasures of the field — a lot of the data is public, and most of the tools are open source. So a new “experiment” for computational biology practitioners can be as easy as clicking a few links, downloading some data (making sure you have enough local storage space — the datasets can be huge), and writing some code. Speaking of which, another recommendation to those interested in the field is this: learn R. This open-source statistical package is an industry standard and my daily workhorse. Through its vast contributor network, R has seemingly a package to do everything, including providing an easy-to-use framework for making web apps for visualizing and sharing data.

So, what does mathematics (at least, the math I spent all that time studying for my PhD) have to do with my new career?  While I’m not proving theorems anymore, I would argue that my PhD experience provided important training for my work in a number of ways. A critical, analytical perspective is obviously important for both endeavors. Also, having a PhD background means mathematics is not a barrier to understanding new statistical techniques, and I can focus instead on the ideas. A love of learning, and a humility and curiosity about what you don’t know, are also crossover values. In my job, as in my PhD study, each day means another opportunity to learn, keeping things fresh and interesting. Finally, this is not a job for those who prefer to work alone. Creating new therapies is a complex, collaborative, multi-disciplinary endeavor, requiring clear communication with all the stakeholders. One of the joys of the position is to work with scientists who are not computationally or mathematically oriented and help translate their questions into concrete analytical problems. Teaching experience in academia has really helped in that regard, since it strengthened my skills of listening and explaining.

For those who love math, love programming, and love learning new things, computational biology is a great career option, and provides an opportunity to make a concrete difference in people’s lives.

1 American Cancer Society, https://www.cancer.org/cancer/non-small-cell-lung-cancer/detection-diagnosis-staging/survival-rates.html

2 “Further Evidence that Immunotherapy Provides a Longterm Survival Benefit for Lung Cancer Patients,” R&D online, 12 Apr 2018, https://www.rdmag.com/news/2018/04/further-evidence-immunotherapy-provides-longterm-survival-benefit-lung-cancer-patients

3 The Cancer Genome Atlas, https://cancergenome.nih.gov/

How I became a Data Scientist

zzlalo Career Information, Data science, Job Seeking, Skill Transfer

BOLOR TURMUNKH Photo

by Bolor Turmunkh, PhD, Data Scientist at Uptake Technologies Inc., Chicago, IL

First Steps

At the beginning of my fifth year of graduate school at the University of Illinois, with thoughts of impending graduation, I started thinking for perhaps the first time in my life about who I wanted to be. I had lived happily as an information hermit for four years. I had spared little thought for anything other than academic research. It would have been handy if I had kept up with career trends, sought-after skills, or internship opportunities. But as they say, the secret of getting ahead is getting started. So, I buckled down and got started.

After a quick google search on trending careers of the future and cross-referencing the required skills with my own past experiences, I landed naturally on data science. In this post, I will recount the path to my current position as a data scientist, and describe some differences between academic research and industry work – so that if you are considering the same options, you might be better informed about the trade-offs.

What is Data Science?

A famous Venn diagram (google “data science Venn diagram”) defines data scientists as having skills at the intersection of coding, statistics, and domain expertise. They are the people who take a business problem, go prospecting for available and attainable data, re-formulate the question in technical terms, design and implement a statistical and machine learning task, and re-interpret the results for the business client to ultimately answer the original question. That makes it sound like to be a data scientist you need to be a statistician and a computer scientist with years of industry specific experience. That’s not quite true.

The reality is, data science is both vast and new, with specializations and sub-fields quickly developing. Highly sought-after data scientists are people who are broadly familiar with all aspects of data science while being experts in one or two fields. It is a highly achievable career for mathematics graduate students – with some preparation.

How did I become a Data Scientist?

There are plenty of resources online that outline possible paths to becoming a data scientist. I will simply describe my own experience.

From the moment I realized I would enjoy being a Data Scientist to finding my first internship, I spent 9 months devouring online and free courses on Machine Learning and Python, sent out dozens of applications, got two interviews, failed miserably at one of them and lucked out with the remaining employer, who was willing to give me a chance. It was a small start-up in San Francisco developing enterprise software in Natural Language Processing. They posted internship positions on their website alone, and I daringly emailed the founder directly with my best pitch about why I would be a good match. I say daringly because it is not standard practice at larger and more established companies. But anything goes at start-ups, and this one happened to go swimmingly for me.

During the three month internship, I learned intensely. My technical knowledge deficit was overwhelming at times. But here, my academic training was an asset. Living with overwhelming stress without it paralyzing you is arguably what “PhD grit” is all about. A harder adjustment was the social aspect of the workplace. Like any profession, there are jargons and topics, popular and unpopular opinions, the latest and meanest blog posts, all exchanged electronically in an open, yet entirely silent office. But I found mentors and allies who helped me feel at home.

This internship was just the beginning of my journey to becoming a data scientist. It took another two years and one failed job search cycle before I landed my current position.

How is being a Data Scientist different from being an academic?

You are no longer alone.

Intellectual isolation was the hardest part of my academic research experience. Apart from conferences in my particular field in mathematics, and research meetings with my advisor, I had no peers with whom to engage in frequent and technical discussions of the details of my work. That is no longer the case in industry. Not only are my coworkers ready to get as nitty-gritty into my project as I wish to go, they also possess a wealth of experience dealing with similar projects and are happy to share their expertise. Learning in such an environment is exponentially faster than learning alone.

Project time frames are shorter.

Time frames differ significantly from company to company. Larger companies tend to tackle longer term projects. Software teams typically have shorter time frames due to the nature of the work. So, the scope and the strategic importance of  your day-to-day activities will vary depending on where you work. For me, deadlines for large projects are on a quarterly basis and smaller ones are weekly. The shorter deadlines are oftentimes helpful since they force you and your managers to clearly define goals and criteria for success. On the other hand, short-term goals can sometimes feel short-sighted, if your team’s priorities change drastically.

You won’t always get to decide what to work on.

This one is a spectrum. Companies such as Apple prefer to set strategic directions and product vision from the top and have them permeate downward. More bottom-up companies such as Facebook prefer a more entrepreneurial feel. Most companies lie somewhere in between, which means you are somewhat in charge of what you get to work on. My team establishes quarterly priorities and project proposals together, which then go through a review process to make sure the proposals align with company goals.

You have more resources.

As a graduate student, the main resource I had was my own time. As such, I was used to solving all my problems on my own. But as a team member, your goal is to arrive at a good solution in the most efficient manner possible. Doing everything yourself is not the most efficient way. Getting help is not only highly recommended, but expected of you.

Done is better than perfect.

It is an entirely new skill for most academics to weigh the costs and benefits of doing the job perfectly versus doing it fast. In industry, one makes this trade-off every day.

Closing thoughts

The qualifications and projects of a data scientist are quite different from those of an academic mathematician, and yet the actual work is quite similar in nature. The great majority of a data scientist’s time is spent defining and re-defining an ambiguous problem until it can be clearly stated, and then solved.

Once a data scientist finds interesting results, it is crucial to communicate them to the end customer. Building a story around a complex issue, supporting that story with evidence derived from data, and interpreting the results into a concrete recommendation for the customer, are the central tasks of the job. From this perspective, your graduate training in mathematics, statistics or operations research will provide a strong foundation for moving into data science.

Good luck with your career transition and job search!

What I know now that I wish I had known then!

zzlalo Career Information, Data science, Skill Transfer

Kristine_Jones

Kristine Jones, PhD – Senior Data and Applied Scientist, Microsoft
Opinions expressed are my own and not necessarily those of my employer or any institution that I may be affiliated with.

When I was first approached about writing this post, I was asked to try to convey what I know now that I wish I had known as I was applying for jobs out of grad school.  My response to questions such as this is always that there is no one course I wish I had taken, no one skill I wish I had acquired, no one opportunity that would have pushed my early career down a dramatically different path. This is not to say that I haven’t reaped the benefits of broad exposure to numerous skills commonly bullet-pointed on tech industry data scientist job descriptions. I would not be doing my due diligence if I pretended otherwise. That being said, hiring decisions for mathematicians based exclusively, or even primarily, on those bullet points are poorly considered. (Learn some coding, stats, and optimization methods, though … it can only help you).

Ultimately, the most valuable skill that a mathematician brings to any team is her ability to abstract the core technical problems the team is facing and provide the basis for solving these problems across all of the scenarios in which they assert themselves. The crux on which this skill rests is an education that teaches students to think deeply and critically about mathematics, both independent of and relative to the breadth of contexts in which it appears.  What follows is a personal history of how I arrived at this view.

From my vantage point as an undergraduate, the University of Chicago embodied this notion of linking independently considered critical thought with the scope of its applications.

The opposite criticism is often levied against the University of Chicago: that it is overly devoted to abstraction in place of the concrete. Perhaps the most telling supporting evidence of this criticism is the slogan “That’s all well and good in practice, but how does it work in theory?”, text often displayed along with the University’s logo on t-shirts sold by student organizations. Other examples abound, even (perhaps especially) from within the mathematics curriculum.  As a senior studying abroad at the University of Chicago Paris Center, I took a course in representation theory.  On the first day, the instructor, trying to assess the background of his students, asked the class who could define a vector space.  Several hands shot up, mine among them, all with the same answer ready – a vector space is a module over a field. The basis for the criticism is not unfounded.

But, while the University does certainly value abstraction, it does not do so without purpose. Even their motto declares this intention: Crescat scientia; vita excolatur.  Let knowledge grow from more to more; and so be human life enriched.

The goal of all the theory, of all the abstraction, of all the critical thought, is to enrich human life.  To make meaningful changes in the way scenarios play out and issues are addressed. To enhance the ways people think and people act.

This idealistic motto filtered down to the mathematics undergraduate curriculum in two ways that were evident to me.

First, mathematical abstraction was not left to the math majors. Proofs and notions of generalizability were at the core of even the most introductory of mathematics coursework, both calculus and non-calculus based. To these students especially, the value to be found in the mathematics coursework was not in a particular formula or tool that they may or may not ever need, nor in seeing jargon from their chosen field strewn out across rote problems.  Instead, they left their math classes with the ability to reason deeply and strongly about quantitative questions they would encounter in the future.

Second, if you wanted a B.S. in mathematics instead of a B.A., you had to take three non-introductory courses in a related minor field. Physics, chemistry, and computer science were common choices, in addition to a more structured option in economics, which was my selection.  This was not for an applied math degree, where the minor field requirement was even larger, but for a degree in pure mathematics. Amidst the modules over the fields, you had to understand what was real about the math you were studying.

I left the University of Chicago not only with lofty examples of abstract concepts and their footprints across mathematical theory, but also with an ability to “suss” out broad connections to this theory across all manner of problems I encountered. I headed off to graduate school to pursue a passion for what I thought was mathematics, with no particular career path in mind. I saw, and still see, a graduate education as an end in itself.

Let’s be honest here, though. Grad school is not for everyone. Grad school is hard.

That’s ok, it’s supposed to be. At its very best, graduate school asks that the student be lost in a sea of seemingly unconnected examples, looking for problems and finding solutions, ultimately weaving everything together into a single theoretical fabric.  It is from this process that graduate education derives its worth. Completing this work is an incredible demand on any person. I was lucky to have a truly amazing graduate advisor. He excelled at guiding his students from examples to theories and back again, all the while allowing them to find their own way.

Which is not to say I was always successful in seeing that path.  I spent one summer doing hundreds of matrix computations only to reach the conclusion that I would need to do thousands more to see if there was any pattern that might indicate the presence of an underlying theorem. My advisor’s comment upon hearing this was that I needed to come up with an “exit strategy for the project,” without saying much in the way of how I might do that. That was the “finding my own way” part.

Moving forward out of that moment was one of the hardest things I have ever done.  It was only when that work was nearly completed that I saw the value of my education was in the struggle to see connections between specificity and generality, regardless of any career decision I would make. More than that, I realized that my passion for math was really a passion for executing this skill. This insight came at a great moment because I was about to graduate, and seemingly out of nowhere, previously unconsidered options abounded.

I sent out resumes and went on interviews.  Microsoft stuck.  Since I’ve been here, I’ve designed large-scale machine learning systems, implemented component-scale algorithms, coordinated projects across research and engineering teams, and performed executive-facing analyses in high-value areas. Very little of this has much relation to the content my thesis (although I did run a lecture series on the theoretical underpinnings of homomorphic encryption that one time), but I doubt I would have been able to accomplish much if I hadn’t gone through the process of writing it. All of my solutions for Microsoft draw their impact from seeing connections between many smaller problems, with many commonalities, and solving them all at once – quite similar to academic mathematics in form, if not in function.

A National Laboratory Internship Experience

zzlalo Career Information, Summer Internship

 

hart_photo
Joey Hart

Joey Hart is a PhD Candidate in the Department of Mathematics, North Carolina State University

I had a very interesting, and on some levels unique, internship experience in the Optimization and Uncertainty Quantification Department at Sandia National Laboratories. The origins of my internship came several months before through collaborations with my eventual mentor on another project Bart van Bloemen Waanders. During the 2016-2017 academic year, I was a graduate research fellow in the SAMSI (Statistics and Applied Mathematical Sciences Institute) program on optimization. Through this fellowship I began collaborations with several applied mathematicians and computational scientists. Our work led to Bart inviting me to apply for an internship under him in the upcoming summer of 2017, thus the story begins.

In the couple months leading up to the summer I was able the talk with Bart on several occasions. He described the project I would be working on in the internship and asked me to do a literature review and propose a plan of work. We were able to iterate with one another to formulate a plan of work which had mutual interest and benefit. I think this was one of the nice and unique features of my experience; many interns receive a plan of work when they begin, I was able to formulate my own before I arrived. At its core, the project was focused on parameter uncertainty in optimization problems constrained by partial differential equations (PDEs).

I learned many useful skills over the course of my twelve week internship; a few highlights of these are finite element analysis, PDE-constrained optimization, C++ programming, and high performance computing (HPC) skills. The learning curve was challenging at first, but incredible rewarding in the later part of the summer. From the onset of my internship, I developed C++ code on top of the Rapid Optimization Library (ROL), a C++ library which, among other things, is useful for PDE-constrained optimization. One of the most important skills I needed during my internship was the ability to read the existing ROL codes and understand how to build on top of them to effectively leverage their capabilities. I began the internship as an amateur C++ programmer and had to learn more C++ while diving into the ROL code. In terms of mathematical skills, having a solid foundation in functional analysis was crucial for me to make rapid progress through the background literature and mathematical formulation of my work. Much of my internship focused on code development because I progressed through the mathematical development quickly. The coding work should not overshadow the mathematical complexity underneath and the important mathematical skills needed before writing the first line of code.

Over the course of my summer internship I learned computing skills which I probably would not have learned at my home university, I gained exposure to applications which have significant societal impacts, and I learned about the organization structure and workflow of a research laboratory. Having an internship at Sandia is beneficial for future staff or postdoctoral applications and it serves as a nice trial run to determine where I may thrive in future positions. Time prohibits me from highlighting the many other educational, professional, and personal benefits of student internships.

To top of the experience, its not over! At the end of the summer I converted to be a year round telecommuting intern. I work part time from my university to continue work on the project and we are moving toward submitting an article for publication. I have discussed future plans with my mentor, but the story is still unfolding. Stay tuned for another post with the remainder of the story!

Mathematicians are Needed in Industry

zzlalo Career Information, Industry ,
Gregory_Coxson
Greg Coxson

At this point in my career, I have worked at a number of organizations, usually technology companies with military contracts. I am convinced that mathematicians strengthen organizations, and sometimes make revolutionary changes, often in small ways that are not celebrated as often as they should.

My first job was at the Center for Naval Analysis in Alexandria, Virginia. CNA performs long-range studies for the Navy, and is one of the oldest military Operations Research firms. I was working for a crusty old radar engineer, who wanted me to perform Monte Carlo analysis of Russian missile raids. This required thousands of runs of the program. One day, I needed to consult with a mathematician on another floor. I was surprised to find that he knew the simulation I was spending my days with. But what really amazed me was when we started discussing specifics. At that point, he pulled out a big binder containing tables of every possible combination of inputs to the model, and the associated outputs. He had invested the time one week to run all the possibilities and compile them. Having done this, he did not need to run the model for hours a day; instead he had just to pull out the binder and find the right row to pull out the results. It impressed me that this approach was much more efficient.

Later in my career, when I was working for another company, we had a large number of engineers working on a new ballistic missile system for the Navy. The schedules were aggressive, and the work multi-faceted and difficult. On one of the projects, it appeared necessary, despite the tight schedules, to spend a year running cases of flight trajectories. However, there was a PhD mathematician working on this, and he argued that since all the factors were known, mathematics could be used to perform a quick study, and come up with all the possible trajectories. He saved the company a year of effort and countless computer runs.

In these cases, Mathematics is not enough. It is important to get the information into the right hands. A junior engineer or mathematician will not be listened to, at least without concerted effort and the right arguments.

I had the opportunity to learn this first-hand. I was working on a critical program for the Air Force, and one evening before heading home, I was reading the specifications (not always easy reading). Before I went too far with this, I came upon something that stopped me in my tracks. Here, in a system where efficiency was highly emphasized, was an operation being done 80 times in one set, and then in the next set, the inverse of those operations was being done. This seemed to me something that should be fixed. So I went to my boss and pointed this out. However, I was new, and my boss did not know enough mathematics to understand my claim that eaeb = ea+b. No matter how I argued, she was not going to take my word for it. Her approach, ultimately, was to arrange a panel discussion with some scary senior analysts around the table to make me retract my story. But I did not back down. Looking back on it, the issue was a badly implemented discrete Fourier transform. I left the company soon after. It took about a month before I started getting phone calls asking for my notes. They had come around to agree with me.

The point of all this is, that mathematicians are needed outside of academia. Mathematics is used, and sometimes misused, every day in almost every industry. Mathematicians are needed for their training, but also their insights. I believe that mathematicians are able to find efficiencies, and new approaches, that others are blind to. Mathematicians are needed to prevent errors, to analyze complex problems and systems. There is no doubt in my mind that we need more mathematicians in industry.

Unsolicited Basketball Coaching Advice from a Student Researcher

zzlalo Sports Analytics, Summer Internship
CharlotteEisenberg
Charlotte Eisenberg

This summer I spent a lot of time with NCAA Division I basketball statistics. As a student researcher at Davidson, I examined at least a hundred statistics tracked across the division, from the height and experience of the players to the team free throw and turnover percentages, each potentially linked to the success of a team. My goal was to find the strongest correlations and use them to predict the outcomes of games. Many of the statistics most strongly correlated with winning games, like offensive and defensive efficiency (points and points against per 100 possessions), felt more descriptive than predictive. I became curious about what sub-statistics contributed to efficiency that could be more easily isolated and coached. What statistics would an analytics minded team focus on to see the greatest increase in their win percentage?

 

In recent years there have been basketball teams (not many) that have staked out their claims to be “the analytics team.” For prime examples look to Belmont in Division I or the Houston Rockets in the NBA. Those teams, and others that are vocal about using analytics, are the teams that think they can’t win on skill alone. The first thing an analytics minded team focuses on is taking more three point shots. The theory behind this approach is pretty solid: three point shots have a lower probability of success but a higher payoff, creating more variability in a team’s scoring outcomes. In a simplified example, lets have Belmont scoring between 60 and 70 points per game going up against North Carolina, which scores 77 to 90 points per game. If Belmont gets no better with their shooting percentage but takes a higher proportion of their shots from behind the three point line, they have the potential to see a wider range of scoring outcomes, say between 50 and 80 points instead of between 60 and 70. Belmont’s chances of winning improve because they now have some small chance of a scoring outcome higher than some of the scoring outcomes North Carolina produces. For every one percentage point that underdog teams increase the proportion of their shots devoted to threes, their winning percentages increase by an average of 2.2 percentage points, according to a Peter Keating article in ESPN Magazine last May. They may take some bigger loses, but they make up for it by eking out some unlikely wins. 

 

If I were coaching an underdog basketball team (or a top basketball team), I would definitely push the threes. One pair of three-pointers equals three two-pointers. Simple. Assuming the opponent is shooting twos, my team’s shooting percentage on threes has to be only 66 percent of whatever they’re hitting. An opponent shooting 50 percent for two can be matched by a 33 percent conversion rate on threes. But I don’t think my team’s success will hinge on shooting percentage or distribution. A basketball team wins a game by scoring the most points, which come from shooting baskets, but getting an edge in shooting percentage is tough to possibly unattainable in DI. In the average NCAA Division I basketball contest last year, the shooting percentages for the two teams were separated by just 3.2 percentage points. Assuming the two teams have an equal number of possessions, the “better” team can expect just a couple extra baskets per game, five or six points in a 70 possession game. When I ran logistic regressions on team stats to predict the outcomes of NCAA basketball games I was somewhat surprised to find that the most impactful statistics (largest regression coefficients) were turnovers, steals, and rebounds, not shooting percentage. The team that creates the most opportunities (possessions) for themselves to score usually scores the most. 

 

My (hypothetical) basketball team would need to understand that possessions are key. The average shooting percentage for a Division I team hovers around 50 percent, so each possession is worth at least one point (depending on the number of shots the team takes from three). If a team can raise their shooting percentage from 50 percent to 55 percent (a huge and likely unattainable success to most coaches), the change would translate to an extra three and a half baskets per game. Every percent point gain in shooting will translate to an extra point or two scored per game, the same result as simply pulling off one more steal. I would classify achieving steals as significantly more attainable than improving shooting percentage. There were 77 Division I teams who averaged over seven steals per game last season, with a benefit equivalent to improving their shooting percentages by at least 5 points. Teaching a team to not turn over the ball is hard. Effective ball handling and passing involves years of skill development, talent, and confidence. On the other hand, increased steals (in my opinion) just take a little nerve and boldness.  I would teach my team to be gutsy, stick their hands out there, lunge for the ball. If my team went for a steal on every possession they would create opportunities for themselves and stun their opponents. It could be a new recipe for success.  

 

An analyst’s dream is to find some underutilized path to success that will allow an underdog team to stun favorites. Teaching a team to emphasize stealing should produce measurable results in scoring efficiency and has the potential to upend the character and mentality of the team. A team that focuses on steals ingrains in its players the value of a possession. Every time a team allows their opponent to dribble down the court, pass and take shots, they are handing over possessions. My team would be relentless on defense, not allowing a single shot or pass or dribble to go uncontested. There is a perception that analytical approaches to sports undermine the value of athleticism by letting teams play smarter instead of harder. My analysis suggests that the smartest thing a team can do is play harder, recognizing the value of every second on the court and not giving any away to an opponent’s possession without a fight. 

The importance and excitement of team science — and how optimization research fits in almost everywhere

zzlalo Career Information ,
XBD201705-00091-010
Juliane Mueller
© 2010 The Regents of the University of California,
through the Lawrence Berkeley National Laboratory

I am currently a research scientist in the Computational Research Division at Lawrence Berkeley National Lab. I got promoted into this position very recently. Before, I was one of the Luis W. Alvarez Fellows in Computing Sciences at Berkeley Lab. 

My education (MSc and PhD) was in applied mathematics. Early on during my Master’s education, I got interested in optimization and I enjoyed learning about different mathematical algorithms that were developed for solving application problems to optimality. The applicability and in particular the usefulness of mathematical algorithms for improving life and solving real life problems intrigued me. After my PhD, I went to Cornell University as a postdoc where I was introduced to optimization applications related to climate model development and the environment. Through my advisor, I learned about the different National Labs and their named postdoctoral fellowships which basically allow a recipient to develop their own research agenda. I applied for and accepted the Alvarez Fellowship at Berkeley Lab. In contrast to other National Labs, Berkeley Lab did not have an optimization research group. Some might consider this as a disadvantage because there is no senior person to give me guidance and feedback on my work. However, the non-existence of an optimization group also poses an opportunity. It means that there are many domain scientists who may have difficult unsolved optimization problems that they may not know how to tackle.  This means a lot of collaboration possibilities, and eventually, with some momentum and a lot of effort, even the possibility to establish an optimization group at Berkeley Lab. 

In contrast to the more academic setting that I had so far been exposed to during my education, Berkeley Lab offered a completely new setting of interdisciplinary work and opportunities to collaborate with domain scientists from all science areas. I reached out to many scientists to discuss about their work and to see if they run into optimization problems that they do not know how to address efficiently. Sure enough, I found many takers. Collaborating with domain scientists and the breadth of application problems I get to work on are the most interesting part of my work. I constantly get to learn about new science areas, I learn new terminology, and I encounter new classes of unsolved optimization problems. Throughout the three years I have spent at Berkeley Lab so far, I have worked together with scientists in climate research, combustion, cosmology, spectroscopy, light source lattice design, and plasma accelerator design. 

I mostly collaborate with scientists who develop simulation models to study physical phenomena. Although the scientists understand the physics extremely well and are able to model the physics with high accuracy, most simulation models have parameters that must be adjusted. This is often done based on the scientists’ knowledge and experience, which is a valid approach for some science areas. But in other areas, an efficient approach for adjusting the simulation model parameters is needed and sophisticated optimization algorithms (and thus my work) enter the game. The challenge lies often first in learning enough about the domain scientists’ work to understand the goals of the research. The next challenge is in formulating a sound optimization problem, and then finally developing new solution methods. The most rewarding part of my work is the excitement of the domain scientists as they are able to use a new tool to solve their problems more efficiently (I make my codes publicly available), as they see completely unexpected solutions that they would have otherwise never expected (which often uncovers characteristics of the problem that were not expected), and as their simulations now make better predictions and allow scientific results to be found more efficiently. 

The reason why I decided to stay at Berkeley Lab beyond my postdoc are the exciting and relevant problems I can collaborate on with other scientists. I constantly learn new things and most importantly, other scientists are completely open to explore new ideas, and they welcome the opportunities to learn about new methods. I have not met a single scientist who wouldn’t agree on a meeting to talk about collaboration possibilities. Obviously, being willing to reach out to and collaborate with other domain scientists is a necessity at any National Lab. Research is not done all by yourself, at least not if your goal is to do more than just publishing. You have to be willing to sometimes go out of your way (meaning your comfort zone of research topics) to explore new ideas. You have to be willing to invest a bit of your time in running some initial optimization trials to see if there is actually any hope at all for the domain scientist’s problem. But this also means that you can use these preliminary results as a starting point for a grant application. 

The soft funding situation (all of your salary comes from projects with finite duration) may make some people anxious at times. Somewhere in the back of your head, there will always be this question of whether or not you will have money next year. That’s a thought one has to be willing to live with, but then again, in today’s world, nothing is certain and I don’t feel too stressed about this (yet). But as mentioned, you need to be willing to approach other scientists, especially if you are a bit junior and not that well known around the lab. I started doing this early on during my postdoc already. Even if not every discussion leads to a project, at least people will know what you are working on and they will come back to you in the future if they encounter a problem they know you can help with. For this, I find it extremely important to be able to talk about my work in layman’s terms since not everyone has a thorough mathematics education. 

My advice for anyone who is looking for a successful career at Berkeley Lab is that you have to be able to work independently as much as collaboratively. You have to be able to come up with novel ideas to solve problems. At the same time, the concept of team science as introduced by E.O. Lawrence in the 1930s remains an integral part of the Lab’s efforts today, and the success of collaborating teams with mixed skills and diverse backgrounds has proven to be the best way to tackle the most complex science problems. Therefore, it is important that you stay curious, that you are open to new ideas, that you are willing to step out of your research comfort zone to learn new things and explore new science areas. The lab setting gives you the opportunity to grow. Take the chance! You don’t have to know all the details of everyone’s research, but keep on learning, keep asking questions in meetings — people are more than happy to explain their research to you. If you are still in college, attend some introductory lectures on topics that are outside your area — engineering, economics, programming. I found that the classes I took on intercultural communication were extremely valuable. National labs attract researchers from all over the world and you will end up in a very diverse setting. Having some kind of an idea how to navigate this setting effectively is extremely helpful. Lastly, be open to talk to people, volunteer to help at outreach and other lab events. It takes some effort, especially if you are more on the shy side, but it pays off, I promise!

Save