|
Magnetic North posted:My boss says that in the next year or so, he wants to start a Machine Learning / AI project. So I've been looking into it in the hopes of being prepared and being able to be a part of it and use it as a means to enhance my career prospects and learn cool stuff. However, while looking into it, there is a lot of linear algebra under the hood, which I never learned. On the one hand, it would arguably be better to really get an understanding of it, even in a relatively short run-up. On the other hand, if we're just going to use some existing open-source library, does it even matter? Understanding LA is a huge plus for any programmer. Anything involving images, representing categorical data, NLP, vectorizing problems will rely on it. Just take a course or two at Khan's and maybe one of Andrew Ng's for reinforcement and applications to plant a seed in your head. IMO it's just one of those things that should be in the toolkit of any developer, especially anyone getting into ML. I can't overstate how important LA has been to me, but I can give you some anecdotes if you're curious.
|
# ¿ Nov 24, 2019 05:58 |
|
|
# ¿ May 22, 2024 17:28 |
|
JIZZ DENOUEMENT posted:When I was in grad school, I really liked my courses on program evaluation and regression analysis. Stuff like STATA. Creds/certs, probably no. I'd probably try to start with driverless AI just to get a feel for what's currently being offered from that perspective, like h2o, data robot, or azure's ML workspace. The latter is rough but it has some promise, with some explanatory functionality in preview mode. Then branch out into tidy looking Jupyter notebooks on a Github somewhere, hopefully focused on real world/business use cases. Get some experience deploying trained models to the cloud and lambda/serverless functions in general.
|
# ¿ Nov 24, 2019 06:04 |
|
mearn posted:I'm currently going through a Master's program in Data Science. It's an online program, which isn't really ideal and honestly I haven't found much of the material to be much more informative than a DataCamp subscription would be. I know an online program isn't really the best choice. I've got two semesters left at this point and it's all being paid for by my current employer, so I'm going to stick it out. I guess my plan from here is to just work on personal projects and try to build a portfolio in the meantime, since I'm not sure how much value this degree is going to have on my resume. Being able to quickly stand up some API is a huge plus for anyone in the field, and some of those folks are awesome programmers. I work as DS lead for a design/prototype studio and the best programmer we have is some guy who loves doing everything in Django. I've learned a lot from him. Learn just enough from that crowd and you can add 'deployed custom ML models to cloud API for real time analytics and what if scenario analysis' to your resume.
|
# ¿ Nov 24, 2019 06:09 |
|
Kim Jong Il posted:
Mostly a good article, though. I think many customers have been sold on ML/AI, but in reality haven't even begun to clean and bring their data together. Add in high expectations of what ML/AI can do and they think the failure is on the practitioners, when in reality they aren't even collecting the right data in the first place. For organizations with an advanced data platform and analytic capability already, the right data scientists will continue to add value. For the rest - there's quite a ways to go. Sure writing Python code and SQL queries is easy, but the low hanging fruit is gone. It's still back to basics for many 'data scientists' - cleaning and engineering data, data lineage, basic visualization, and quickly prototyping ML/AI applications with the latest cloud technologies.
|
# ¿ Nov 24, 2019 06:59 |