Joining the Lab
To be frank, we’re rarely 100% certain our lab will be accepting new students in any given year. It often depends, as much often does, on funding; if funding is looking good we’ll probably be looking for more students. However we’re always happy to have students rotate in my lab, should they be accepted, regardless of the funding situation.
The barrier is never lack of research projects.
If you are interested in applying as a PhD student, note that Dr. Voytek accepts students from either the Cognitive Science or Neurosciences PhD programs (you can find application information for each here: Cognitive Science; Neuroscience). In neither case are students accepted to join a lab, but rather they are accepted by an admissions committee to join the program to which they applied. During the first year of training, PhD students are expected to rotate in 2-3 different labs to gain some breadth of experience and get to know their prospective PhD mentor. At the end of the first year, students then speak with their prospective mentor to officially join a lab.
Do not apply to these PhD programs simply because UCSD is prestigious. That’s not to say that prestige is wrong, but it’s not the reason to get a PhD. If we get the impression that you’re simply chasing prestige, we will likely not accept you into the lab.
The absolute best place to pursue a graduate degree is the place that maximizes the following variables (weighted according to your personal taste):
- Does the city “feel” good to you?
- Will you be happy living there for the next n years, during times of frustration and hardship?
- Are the faculty going to support and challenge you?
- Are the faculty sufficiently knowledgeable about the field in which you are interested?
- Does the university support you financially well enough given the local cost-of-living?
Factor in significant other and/or family happiness!
- Does the department have an active community (social, academic, and otherwise)?
- What is the teaching load expected of the students?
- What is the course load?
- Are there opportunities for mentoring undergraduates (who can also help with your research)?
Prioritize academic interests and happiness. You can mostly ignore university “prestige” as it seems to be only weakly correlated with faculty utility in any given domain. Instead focus on the strength of the faculty, post-docs, and students.
In short: if you’re not happy, your work will suffer for it, and frankly, prestige isn’t going to make you laugh, go to lunch with you, or carry you through your PhD.
With regards to specific skills/knowledge before entering a lab, we personally believe that the most bang for your buck would come from learning to code. Our lab is Python-heavy, so we recommend learning or brushing up on your Python skills at Codecademy and iPython.
Also, nothing beats good statistical knowledge. Specifically, understanding the GLM will carry you far. The rest you can learn during your PhD (that’s what the PhD years are for: learning!).
A lot of people ask Dr. Voytek about his industry time working as a data scientist, and what they need to do to break into the field. When I was on the hiring side, I always looked to see some portfolio showing evidence that the applicant had done something. So many people would approach me with a resume and say, “here are the classes I took,” but, honestly, so did tens of thousands of other people so why should we hire you?
Thus my advice is to do something.
brainSCANr was a huge help for my data science career for that reason: it was clear evidence that showed we could conceive of an idea and see it through to completion. Even something silly like this analysis and visualization of the “best rap vocabulary” shows that the author can think of a neat idea, run a fun analysis, complete the project, write it up, and visualize it.
That’s what some companies are looking for. If you need inspiration and/or practice, check out kaggle.
As for my data science toolkit:
- Anaconda to manage python and packages
- iPython/Jupyter Notebook for quick scripting
- Github for version control
- matplotlib + numpy + scipy + pandas + scikit-learn + statsmodels + nltk + BeautifulSoup + bokeh for statistics and data visualization
R is fine, but clunky (in Dr. Voytek’s opinion), and Python plays with web applications and allows for interactive data visualization (via bokeh) much better.
Also, Python is used extensively for non-stats/non-scientific programming, so if my students learn Python they have a marketable skill should they decide to not pursue a career in academia.