These are chat archives for FreeCodeCamp/DataScience

28th
Sep 2016
Eric Leung
@erictleung
Sep 28 2016 01:26
Here's perspective about "The Fundamental Limits of Machine Learning" I found interesting http://nautil.us/blog/the-fundamental-limits-of-machine-learning
evaristoc
@evaristoc
Sep 28 2016 12:04
@erictleung absolutely... :+1:
Hèlen Grives
@mesmoiron
Sep 28 2016 12:58
@erictleung thank you for the article. I made a comment on it with a wink: That sounds familiar. Do engineers have found their artificial children? Raising and educating humans is the go to algorithm of choice to produce highly sophisticated learners. From that perspective we have chosen the the most slow, costly way to reproduce what nature already had figured out and the brain did evolve for a reason or natural law. Deep learning can be quite useful and nothing wrong with it as long as one does not mistake the map for the landscape. That is something that is easily done if we trust too much on what is discovered. Hence things like over medication because complexity is blurred. Drawing lines are choices too. Humans are great at pushing other humans in unnatural positions that only happen to be harmful over the long run. How long have engineers been working on deep learning. How old is their child now?
CamperBot
@camperbot
Sep 28 2016 12:58
mesmoiron sends brownie points to @erictleung :sparkles: :thumbsup: :sparkles:
:cookie: 414 | @erictleung |http://www.freecodecamp.com/erictleung
evaristoc
@evaristoc
Sep 28 2016 16:04
@mesmoiron Now and then I must admit I get blind (not an engineer, just interest in tech). Nice there are people like the guy of the blog who bring us back to Earth. It is not about underestimating the power of the technological development but also as important not to overrate it.
Eric Leung
@erictleung
Sep 28 2016 17:58

@mesmoiron @evaristoc coincidentally this short article published just 5 hours ago discusses similar things.

But there are important limits to machine learning, and the biggest of these is that it still requires humans to frame the right question.

But good news for those who know machine learning is that

Mahaffey, however, cautioned that while deep learning represents a remarkable technology, many firms still need to learn the basics of machine learning.

And as with all data analysis, it's always good to keep in mind

Simply shoveling random stacks of information, he said, will produce a “garbage in, garbage out” result.

Eric Leung
@erictleung
Sep 28 2016 18:06
Here's another article on 14 Traits of the Best Data Scientists. I like trait number 13: "They're Human" :smile: I think you could argue that these traits are more universal in finding the best <INSERT JOB TITLE>. Good things to consider nonetheless.
John L.
@jomcode
Sep 28 2016 18:24
has anyone used f# for data science?
evaristoc
@evaristoc
Sep 28 2016 19:45
@erictleung I can't be more agree with those... unfortunately it might be rooted in some organisations that you think (I was working for a market research company, so I have seen things... :) ).
@jomcode not me... seems to be closely related to .NET and C# too... haven't tried that path either. Interesting your choice. Any reason for that?
John L.
@jomcode
Sep 28 2016 19:48
its a functional language and i like all of the new dotnet core stuff. just wondering if it'll be a hindrance to use when it comes to this. i'll set up a dev environment for another language if i have to
evaristoc
@evaristoc
Sep 28 2016 19:51
It seems an interesting choice, I was checking on Internet and found it could be of use in the Financial sector if that is your path. To be honest, haven't heard much about F# in the circles I am moving around, but it looks like a very powerful language indeed. Not the easiest to learn though. If you are into functional I should say try Scala? Already popular and shares some stuff with F#, but the haskell-like syntax :) .
John L.
@jomcode
Sep 28 2016 19:52
are you saying scala shares a haskell-like syntax or does not?
ive only looked a snippets from scala never really looked into the language
evaristoc
@evaristoc
Sep 28 2016 19:53
In my opinion and for what I think I know about haskell, less than F#.
John L.
@jomcode
Sep 28 2016 19:54
ah i see. my favorite thing about f# is the way the types work.
evaristoc
@evaristoc
Sep 28 2016 19:54
What exactly?
Don't know, asking you...
Checking some quick code... do you mean the type statement? Looks a bit like a class but MUCH cleaner...
Good luck with that!
John L.
@jomcode
Sep 28 2016 19:58
yeah exactly. the types are kind of like constructors
evaristoc
@evaristoc
Sep 28 2016 19:58
yea.
indeed...
John L.
@jomcode
Sep 28 2016 19:59
idk i just prefer it probably because im just getting into using a functional language and im still trying to break free from the oop way of thinking :)
evaristoc
@evaristoc
Sep 28 2016 20:01
:) :) :) That is an internal battle we are all having here. At least the majority... I am sure there have been more than one casualty and break downs

I will check, but for now... hmmm... F# is more in my waiting list... thanks for bringing it on!

Let us know about your projects, if you have something in ML with it? Show your code! Having to go. Take care!

John L.
@jomcode
Sep 28 2016 20:06
alright have a nice day