How to Do Great Research A Discussion Chand

How to Do Great Research A Discussion Chand T. John

What’s going to happen in this lab meeting?

What are we going to do? • • Break down research execution into goals Identify challenges in reaching goals Discuss how to progress despite challenges Summarize results of our discussions AND… Be entertained Relate to each other Motivate future discussions and talks

Why are we doing this? • I really want us to do this • Scott really wants us to do this • You really want us to do this, because – These issues affect your life – You’re not alone in dealing with all this – Discussing issues is better than silence • This discussion is waiting to happen

How do you do great research?

Scott says… The overall goal is to engage in research that is unconventional, surprising, and excellent that can change the world and make it a better place. This will be more fun and open doors for your future.

To do that, Scott says… • Know everything. This is the key to immerse yourself and know all the literature. • Be creative. Think about what problems will really make a difference versus being small incremental contributions. • Be resourceful. Use all the resources at Stanford and around the world to gain experience and help define and execute your project. • Set goals. • Pursue goals with speed and vigor. • Find passion. Starting with the goal of life balance generally leads to mediocrity. Finding passion and pursuing your vision with energy leads to more passion and more energy. Some people have passion in a relatively narrow area; others pursue lots of different areas of human endeavour with passion. All of this is awesome.

Felix says… 1. 2. 3. 4. Select a good topic a. b. Interest to the scholarly community at large Mainstream vs. groundbreaking advantages/disadvantages Methodologically flawless a. b. Mentor review Peer/colleague review Executable a. b. c. d. Investigators environs Lab & monetary support Time needed Personal commitment Publishable a. b. c. Personal examples (strengths & weaknesses wrt 1 to 4 above) 1. My thesis 2. Motor unit research at NIH 3. Motor unit research at Univ. Maryland 4. Simulation work a. Jumping b. Pedaling c. Walking Others’ not executing the same research Journal quality Quality vs. quantity of publications

Any comments? • Cat gastroc project: satisfied all criteria except peer/mentor review, and was scooped by a group in England; only got an abstract, no paper • Jumping had low impact on first glance, so had to sell it as a way to eventually get to walking • Pedaling had higher impact • Now doing walking, and Felix retired

How do you actually do this?

The Reality 4/11/2007 http: //www. phdcomics. com/comics. php? n=848

How do we get… from here… to here? Know everything. Be creative. Be resourceful. Set goals. Pursue goals with speed and vigor. Find passion.

Challenges and Discussions

General Challenges • Losing courage (Ivan Sutherland) • Unfamiliarity with topic, society, environment, expectations of you • Pressure to produce, perform, prove yourself to collaborators, peers, professors

Know everything. • You’ll never know everything that is known. • Unstructured learning: how do you explore what you don’t know, systematically and efficiently? • How do you read regularly while under pressure to produce research? • How do you identify important papers in a field? (MIT AI lab page) Pat Hanrahan

How do you explore a problem? • You're a 3 rd-year EE Ph. D student at Stanford. • You just passed quals and are going to start a project in robotics that requires designing a new controller for a quadruped robot. • You've never taken robotics or dynamics, but you've taken a basic course in control and a course in linear algebra. • In one quarter, your professor expects you to: – Do a literature search on control of quadruped robots and related topics. – Implement a simple PD controller into a simulation framework created by a former student in the lab, for which there is no documentation. • How do you proceed?

• Clarify problem: very specific • Lit search – Groups/names; start on websites – Critical reading/shortcomings->research – (Peers/mentors) • Start with simple stuff end to end

Be creative. • You either have it or you don’t? • How do you productively search for ideas, instead of just letting your mind wander? Or is the wandering mind the creative one? ! • Do you have to know everything before you can pick a cool problem? • Do you play to your strengths or do you go outside your comfort zone?

• Peer/mentor advice • Felix: sitting around thinking generally not useful • Strengths vs. change is personal preference • Slow to change as student since start-up time is high • Experience makes everything easier, including start-up time

Be resourceful. • • • Being isolated/ignored/individualistic Purism Not-invented-here syndrome Pressure to produce original work, not share ideas Taking advice literally vs. interpreting it and understanding intentions • Asking for help vs. doing it yourself – Asking for info is free? Actual work help is free when small, or gets you acknowledgement or authorship? Or collaboration! Or do you simply get used?

• Felix: best students asked for a LOT of help; they picked a good problem and realized they didn’t have the resources to do all of it, so they brought other people in, as collaborators and co-authors • Jeff/Ajay: double-edged sword; have to develop skill of figuring stuff out yourself instead of asking for help a lot, but also have to not waste time on re-inventing the wheel

Individualism vs. interaction • You are someone who generally doesn’t like asking for help. • You work on motion planning for a robot that could one day travel on Mars. • No one in your lab knows much about your work. • You’ve tried about 15 different motion planning strategies for the robot. • Some methods worked okay, and some were horrible. • What do you do? – – Do you try to publish what you’ve done? Do you ask for help or attention from outside? Do you read more papers? Do you keep on truckin’?

Set goals. • Planning vs. actually working – Don’t worry, just keep doing the work, and as time goes on things work themselves out. • Planning vs. trusting your instincts • Travel light. Ron Fedkiw (Keeping a lot in your head and trying lots of things vs. writing stuff down and focusing narrowly on really tiny tasks at a time) Alan Cline

• Plan first to get goals determined • Write down obstacles/challenges as you reach them, even if moving on after that • Me: let self be buried in problem for a while to make progress on a single step; plan to figure out what overall steps (goals/subgoals) actually are

Work vs. sociology, planning • • • You are preparing for your ME quals. You have to review your 5 areas: control, linear algebra, dynamics, biomechanics, and robotics. You’ve taken a course in each area, but you can’t remember much from control, linear algebra, or robotics. Your quals committee consists of hard examiners for control and linear algebra. Your quals talk is taking a long time to prepare: • How do you proceed? • • – Its organization is poor. – It’s not visually appealing. – You’re being advised to avoid certain aspects of your research to avoid upsetting any of your committee members during your talk. You think this is cheap and underhanded, and a waste of your time when you should be focusing on actually studying material and working on other aspects of your talk.

Pursue goals with speed and vigor. • Abstraction, general/global thinking vs. rote work and getting buried in details • Emerging from the details • Focusing on work vs. sociology (how to get published, networking, what to say or not say because of so-and-so…) – Working on the image others have of you – Time selling/presenting >= research work time • Your strengths and weaknesses vs. what’s expected/required of you • Pros/cons of your natural psychology (obsessive behavior, high expectations, purism, self-doubt or impostor syndrome) • Jumping to the end vs. careful, methodical, thorough investigation (do you have to be an expert researcher to be fast and still make good decisions on what to work on? ) • Can you do high-quality independent research as a student of research?

• Felix: selling research helps you learn what you know, don’t know • Jump to the end: a way to plan, visualize the end goal and see what the value/impact is • Felix: actually sacrificing some carefulness in trying to jump to the end goal: okay, e. g. in modeling, we know it’s not a perfect model anyway, but still some value in imperfect information; e. g. for walking, need very little data to know that walking involves one leg taking a step while the other leg is weight-bearing

Jump to the end vs. careful • • Threat of getting scooped => speed! Funding/deadline Conference deadline Graduating

Be passionate. • Passion vs. what needs to get done • Getting pulled in multiple directions • Being motivated to push hard vs. being patient amid obstacles, distractions, setbacks, extreme slowness of progress, rarely seeing positive results of your efforts • Too much passion: perfectionism • Unwillingness to take small steps • Being a research machine vs. being human (eat, sleep, make mistakes repeatedly, lose track of things, have emotions); compartmentalization is key! • Being balanced and caring about society vs. living in your Stanford bubble and pretending to care about society • Passion? Or obsession?

Passion vs. patience, steps • Putting everything into your work, not paying attention to other things • Being patient and taking small steps when you really badly want to reach the end goal • Patience with obstacles that slow you down even further

Future discussion/talk ideas • • • Professors, students, others “How I got where I am” series The computational method More of the same type of discussion Talk on Sutherland, Hamming advice Other ideas?

Other References • How To Do Research In the MIT AI Lab http: //www. cs. indiana. edu/mit. research. how. to. html • The Ph. D Experience, by Mihir Bellare http: //www. cs. ucsd. edu/~mihir/phd. html • Ph. D Life http: //www. findaphd. com/students/life. asp

Thanks, everyone!