Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: What journals and blogs should I be reading to become a data scientist?

news.ycombinator.com

Post reply on HN
Print view View original

21–30 of 46 posts

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#21

Post by nashequilibrium »

Follow the link below, there is like 24hrs of lectures, including materials, code etc. These lectures cover reading data, saving data, cleaning & reshaping, visualization, stats, 8hrs machine learning in scikit learn, version control & unit testing, geospatial analyses. This is all in python using numpy,scipy,ipython,pandas and scikit learn as the base tools. You will love the ipython notebook! https://conference.scipy.org/scipy2013/tutorials_schedule.ph...

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#22

Post by stephencanon »

mswen wrote: post #5

Becoming a data scientist isn't a matter of reading journals and blogs. You can get a sense of the field and what is required by reading those sites but becoming a data scientist is years of hard work. You need to develop serious skills in at least 4 of the following disciplines. Statistical analysis RDMS query development NoSQL databases Machine learning Natural Language Processing Web crawling and data harvesting t…

Well that’s just, like, your opinion, man.

I’m not a “data scientist” (or statistician, for that matter), but of the (excellent) data scientists that I know, the only specific skill they really have in common is statistical analysis. I’d say the truth is probably closer to “statistical analysis + ability to do independent research + computational chops using whatever their tools of choice may be"

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#24

Post by allochthon »

I don't have a PhD, and I'd love to be called a "scientist." But I think it's pretentious to use the label "data scientist" for anyone with solid stats experience and a gift for exploring data. To my mind, scientists have gone through formal training and earned a PhD, which, in a given context, may or may not be necessary for what these guys are doing.

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#25

Post by mswen »

stephencanon wrote: post #22
mswen wrote: post #5

Becoming a data scientist isn't a matter of reading journals and blogs. You can get a sense of the field and what is required by reading those sites but becoming a data scientist is years of hard work. You need to develop serious skills in at least 4 of the following disciplines. Statistical analysis RDMS query development NoSQL databases Machine learning Natural Language Processing Web crawling and data harvesting t…

Well that’s just, like, your opinion, man. I’m not a “data scientist” (or statistician, for that matter), but of the (excellent) data scientists that I know, the only specific skill they really have in common is statistical analysis. I’d say the truth is probably closer to “statistical analysis + ability to do independent research + computational chops using whatever their tools of choice may be"

Agreed it is just my opinion. And rarely, if ever, will you find all of these skills in the same person. More often it will be a small team of people each with 1 or 2 specialties plus some other areas they are reasonably competent at.

Most of what I have been reading on the topic seems to define data science as the intersection of the kinds of things I have listed. I guess my larger point was that each of these areas have their own learning curve and some like statistics or machine learning benefit from formal training. A person does not become a data scientist by reading blogs and journals.

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#28

Post by MrMan »

Unless you are part of a vanishingly small group of autodidacts who can train themselves up to graduate school levels of expertise in multiple overlapping subjects - statistics, computer science (might be able to get away with just being an ok programmer), and the interdisciplinary combination of those called "machine learning," you should disappear into a statistics degree program, and amend the traditional stats program deficiencies with the modern-day leavening agents that create "machine learning."

Downloading scikitlearn and R and such is not going to work. At that level you are only qualified to be bossed around by a real scientist or statistician. You are an "analyst".

Re: Ask HN: What journals and blogs should I be reading to become a data scientist?

#29

Post by kriro »

joshvm wrote: post #20

Don't bother with journals - in pretty much any subject - unless you have a degree and/or you understand what to look for, or are directed to notable articles in bibliographies or by peers. There is a lot of crap in all journals, it's often needlessly technical for practical purposes or too bleeding edge to actually be useful yet. I'm not trying to be snarky, but honestly unless you know what you're looking for it's…

While a little off topic I think my typical research process goes like this:

- Hear or read about something that sounds neat

- See if there's a wikipedia article (I always cringe when I hear some colleagues of mine say never, ever use it)

- Get a high level understanding of the topic from the wikipedia article...that usually leads to some other wikipedia articles + plain old Google searches...just fishing for whatever comes up [I also search for TED talks, youtube videos and MOOCs related to the topic]

- Scribble stuff down on a piece of paper and structure it in a way that makes sense to me (sometimes it's just a list, sometimes a full blown mindmap) ...at this point I have a decent high level understanding...which basically means I could describe the topic to someone without stumbling (which I usually try at this point)

- From the high level understanding I usually also get: key terms for searches, intor level books/articles that are linked etc.

- At this point working at a university comes in handy because it lets me be behind the annoying paywalls at will...search Google Scholar or similar databases for the mined key words. Everything that looks remotely interesting...oh wait BEST TOOL EVER

- Zotero is sick good, comes as a FF plugin...great. If you search in scientific databases and the like a little icon pops up in the address bar of the browser indicating it identified the sources...click, mark everything -> it goes into your collection (with full text access) [I order it by topic so for AI I might have Expert Systems and Rule Based, Fuzzy etc.]

- So basically I just wade through the databases and get everything that sounds interesting from the title into Zotero. Alays a good idea to get some "history of XYZ" or "XCY since author Y" sources

- Once done I read the abstracts and the conclusions and put a rough note what the articels are about. I also scan the sources to grow my collection of relevant articles (I mark what I don't think is relevant or move it into a special subcollection)

- I usually try to establish a history of the field with the major stepping stones, this is usually easy (sometimes not, worth a paper to make it easier for future researchers :P)

- If it's related to programming in any way I also search google or github directly for anything related. Code is good :)

[often there are tomes that are the de facto standards in their fields that serve as a massive source collection as well. Perfect example would be AI - A Modern Approach]

Post reply on HN