Hacker News Reader

Comment by martingoodson | Hacker News Reader

Author here - in case of any criticisms or comments.

A lot of data scientists these days (me included) are former academics with backgrounds in numerical simulation in fields like chemistry, physics, mechanical engineering etc.

They live and breath numerical linear algebra and are comfortable reading advanced theoretical books or papers.

It's easy for them to pick up the basics needed to pass interviews and find a data science job. How would they go about adding some rigor to their understanding of ML and statistics?

I personally love this for statistical inference: http://www.springer.com/cn/book/9780387402727

and this for statistical ML: http://statweb.stanford.edu/~tibs/ElemStatLearn/

I wouldn't expect the majority of data science jobs to be particularly focused on the math behind the algorithms. Rudimentary understanding of probability and how to translate the jargon into your academic background's jargon is more important than deep understanding for these jobs. Passing the interviews for these jobs is one thing. Unless you're specifically looking for jobs that focus on generating new modeling techniques or algorithms for computational statistics, expect to be far removed from even basic linear algebra in actual practice. Source: me. I fall in your described bucket and have worked in data science/machine learning jobs in both contexts (new modeling techniques/stats versus application of off-the-shelf tools).

Thanks for the excellent article.

One more interesting thing I have observed in data projects failing: organizations culture around data and the gap between data science team and engineers. Say, you have 2 top notch data scientist who know enough (stats, markov chains, algorithms and so on..). But let us say an average engineer in the organization doesn't know even a bit about A/B testing or difference between building a machine learning model Vs. obtaining predictions from already built model. Then no matter how good your so called data scientist are, the end result in terms of product or solution delivery is always sub-optimal. If the engineers and data science teams can't speak a common language, the result is always disastrous. Note that the gap is specifically about understanding data analysis as a domain.

The efforts to narrow down this gap must be driven by the lead data science member or CTO. Something like 'data bootcamp' mandatory for every new joinee can help. I had read about Facebook having such a bootcamp mandatory.

I enjoyed the conciseness of the article, but I don't have a big stake in this field so I can't comment on the content itself. I wanted to make a small formatting note: having key parts of the article only as images (not reproduced or described in the text) is bad for accessibility, search engines, etc. an article should still make sense if you take out all the images

I think point 7) needs work. Often times people use words like interpretable to avoid having to think about the data - usually in the context of linear or logistic regression. The model seems "interpretable" because the coefficients are "meaningful" - but often times the model is just as much a black box as other models, for instance regression coefficients depend on what other features you include and the scale of those features. Similarly regression p-values are very easy to misinterpret. I think you should use the data to determine what the model is doing regardless of the model you use.

In summary, it is important to iterate quickly and to validate your results. Using complex models, like gradient boosted decision trees, can often iterate much more quickly than simple models because you don't have to do extensive data preparation. Many analysts are stuck in the mode of using linear or logistic regression for every problem, when there are better tools out there.

I absolutely love this list and have bookmarked it to send to the software director next time he talks about "monetizing" our data.

I think this is great, but point #4 is way off the mark. It's pretentious and it's gatekeeping. It's not true that a strong analytics leader as you've described is incapable of understanding selection bias, measurement bias, Simpson's paradox, and statistical significance. I would posit that a great analytics leader needs to understand all of those. It's not true that only a "data scientist" with a bunch of fancy programming and modeling experience is the only one capable of understanding bias.

Also you misspelled "breathe". As in "live and breathe".

Thanks - appreciate the comment. I guess the point is that anyone who fully understands these things is a 'scientist', regardless of job title.

Reply on news.ycombinator.com