These are chat archives for FreeCodeCamp/DataScience
discussion on how we can use statistical methods to measure and improve the efficacy of http://freeCodeCamp.com
@QuincyLarson @becausealice2 that picture was possibly a good case when not to use a pie chart. In general, I would use pie charts in cases where either the differences are visually huge, at least when compared to the target segment under analysis, or when those differences shown in the chart above are considered statistically insignificant.
I wouldn't exclude them just because. IMO that would be a sort of fundamentalism I personally think scare good solutions.
Perfection is the enemy of the Good.
The following is an EXCELLENT work by Sacha Greif, really. But to see they used a nested bubble charts instead of pie chart just because pie charts are vetoed is very very very but very very very very but very very very laughable:
Really... very very laughable...
@evaristoc @becausealice2 3 months is quite fast. If you really are putting 2-3 hours a day consistently, that is quite a bit of time (i.e. 180-270 hours!), though. I wouldn't say you could be an expert at machine learning, but I think with that time, you should at least be familiar with some core concepts and popular machine learning algorithms. Google even created a machine learning crash course that has 15 hours of content.
With 3 months of dedicated time, I think you will at least have set yourself up to more comfortably learn (i.e. struggling but have the confidence you can eventually learn it) about more difficult concepts and ideas. But yes, as @evaristoc has mentioned, I've also been thinking about machine learning for a bit and still don't think I'm there either :laughing:
@erictleung @becausealice2 excellent points. Tbh. I’m completely self-taught programming. I’ve done extensive networking (through networking, meetups, conferences, etc.) to add something to my isolated learning.
Guidance from people on the field can help save hours upon hours of unnecessary frustration. We kind of get this indirectly when we use someone else’s package, or read someone’s book. I think environment factor can also help curb imposter syndrome. You can really dig into what knowledge is common or rare in the industry, and see where you stand if you ask beyond surface level questions and for advice. Becoming an expert in something is probably a combination of all of these things (isolated, consistent learning, networking, practical experience, theory, relationships [mentorship and sponsorship]).
@becausealice2 @Goldberg @erictleung I totally agree with the need of mentoring. If I am not following that way it is because it is expensive, not because I don't think it is not relevant.
The cons of following this route is that you might end applying solutions that are not good enough. You have a larger chance of adopting bad practices by trying to solve everything on your own, and failing to pass the "no reinventing the wheel" test.
However, there is a huge benefit on doing it on your own and it is that you end up developing your trouble-shooting skills. You are also forced to try different solutions and experimenting a lot.
I like the peers formula though. A lot. My ideal for this room was a sort of collective or similar. No much of that I am afraid. Hard when everyone is after different agendas and responsibilities.
I still hope for a better future though :) . Not that I don't enjoy people like you sharing stuff, comments and discussions!
For that, I thank you :)
evaristoc sends brownie points to @becausealice2 and @goldberg and @erictleung :sparkles: :thumbsup: :sparkles:
Below a simple approach to text analysis by combining classical reading with some IT help.
Found that I can easily spot some main ideas from short texts by simply following the most frequent words. Some classical reading and manual input is still required, but detecting key phrases using this simple technique is not only easier but also help topic detection.