These are chat archives for FreeCodeCamp/DataScience

20th
Oct 2016
Alice Jiang
@becausealice2
Oct 20 2016 00:06
I came to the first meetup for that group that was going to teach a bunch of newbies with no experience at all R and they've changed to the Microsoft DS curriculum which I'm already about halfway through So I ended up just being a source of information for people who have questions about statistics or project ideas :/
I feel semi-useful but also like I didn't really need to come here tonight
Alice Jiang
@becausealice2
Oct 20 2016 00:34
I just recruited at least one new person to come in here to ask questions. I'm a bamf :sunglasses:
Artur Arsalanov
@kurumkan
Oct 20 2016 03:55
@erictleung thanks! I learnt physics. I like math=) Yes I know some statistics.
CamperBot
@camperbot
Oct 20 2016 03:55
:cookie: 420 | @erictleung |http://www.freecodecamp.com/erictleung
kurumkan sends brownie points to @erictleung :sparkles: :thumbsup: :sparkles:
Artur Arsalanov
@kurumkan
Oct 20 2016 03:57
I speak python
Hèlen Grives
@mesmoiron
Oct 20 2016 08:12
@kurumkan I posted a while back some resources. I'm doing now Harvard CS109; I like it. There are enough totally newbies in the room; so they teach very clearly. It's python based. However the resources are a bit difficult to get due to old flash player and often access denied problems. But as long as it works; it's great resource.
Artur Arsalanov
@kurumkan
Oct 20 2016 08:21
@mesmoiron thank you! is it online? what is the cost?
CamperBot
@camperbot
Oct 20 2016 08:21
kurumkan sends brownie points to @mesmoiron :sparkles: :thumbsup: :sparkles:
:cookie: 278 | @mesmoiron |http://www.freecodecamp.com/mesmoiron
Joseph
@revisualize
Oct 20 2016 15:34

Greetings DataScience Group. Last night I attended an event at CodeFellows.
Join us for a special Python seminar, where attendees will create a data project in Python!
During this class, we’ll run through the process of taking a data project from conception to data cleaning and reduction, all the way through to creating a visualization. Along the way, we’ll step through some basic arithmetic operations, justifying our choices with the goal of our study and the limitations of our tools.
This seminar will focus on introducing and immediately using tools in Python.
Check out the full outline for the evening: https://codefellows.github.io/data-analysis-glance

It's a pretty good walk-through to get your feet wet in a little data science-y stuff.

Amelia
@apottr
Oct 20 2016 15:45
This looks cool, thanks @revisualize !
CamperBot
@camperbot
Oct 20 2016 15:45
apottr sends brownie points to @revisualize :sparkles: :thumbsup: :sparkles:
:star2: 2066 | @revisualize |http://www.freecodecamp.com/revisualize
evaristoc
@evaristoc
Oct 20 2016 19:22
@revisualize thanks! Worth checking!
CamperBot
@camperbot
Oct 20 2016 19:22
evaristoc sends brownie points to @revisualize :sparkles: :thumbsup: :sparkles:
:star2: 2067 | @revisualize |http://www.freecodecamp.com/revisualize
evaristoc
@evaristoc
Oct 20 2016 20:04

People

Meetups today: Social Responsible Data Science

In this opportunity I attended this meetup:
https://www.meetup.com/Amsterdam-Data-Science/events/232697475/

Both presentations were interested but perhaps the most interesting one of both was the last one, by a Statistician Prof. at Leiden University, who is also a phylosophist. He reminded the audience about how doing BAD statistics can take us to non-human centred decisions, but biased ones. We have discussed the problem here in this channel, specifically when referring the apparently bias of selecting over higher North American black people recidivism ranking based on machine learning algorithms, which was also used as example in this presentation. Interestingly, there is still no simple answer by people at the level of that professor about how to face that bias. He, as a humanist, suggests though that the best way to avoid the inhuman selection is by simply taking any sensible demographic out of the model. I think this is a fair solution. Here another pair of links about the same topic of last week:
https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing
https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm


In the same meeting in an after-thought and chat I met a couple of people who are working on interesting projects in Data Science for Universities. One of them is working on a search engine for Data Science trainings between Dutch universities. It sounds simple but the actual technical implementation is tough as the material to be used for the search engine is not in a centralised database but scattered in many different places and formats.

The other person was on an start-up working on a (AI?) platform for education staff fully oriented to learn Math and Science for university students. The software is to help teachers to verify the solution of the problems students post as well as providing them advice about better ways to solve it. It is just like an assistant teacher that shows the required theory when needed but forces to iterate with several solutions until getting the right one. The system "reads" your solution and provides the best advice to your case. I tried the demo: no bad... The key offer of this cloud based tools consists in the following:

Correcting homework and (practice) tests is not necessary anymore (sic). Intelligent feedback helps students where they need it. Our solution provides real personalized learning, even with large groups.

Also:

A MOOC mostly contains theory, for example as video-lectures. The strength of our solution lies in our interactive elements, like exercises and tests, which offer immediate personalized feedback where needed. This is something a traditional MOOC can never offer.

It started 2010, so I must say it looks a matured stuff. The format also reminds me another learn-to-program platform developed here in the Netherlands for K-12 classes and that it has been recently deployed.

SoWiSo: have a look if you are just interested:
https://sowiso.nl/en/index.html

evaristoc
@evaristoc
Oct 20 2016 20:09
Regarding SoWiSo above: I have already mentioned this sort of incoming software technology here in this channel, by the way.
Hèlen Grives
@mesmoiron
Oct 20 2016 20:17
@evaristoc are you Dutch? Nice what's your background? University or research affiliated? Don't want to be rude (typing the ipad is tedious in the dark) just very interested about the Dutch landscape
Hèlen Grives
@mesmoiron
Oct 20 2016 20:30
The problem with bias is that everybody who can code and do data science or machine learning can write biased algorithms. Than academia are too much in an ivory tower. They better focus on guidelines; checklists or simple examples that attempt to avoid bias. There sre so many risk situations where such biases may occur. An effort to avoid is like teaching good reasoning skills. Especially with insufficient math background people use off the shelf tools. Often output is blindly believed. The computer made the mistake is so widespread. Precisely because of the mathematification the effort should lie in human understandability of processes. Why does an alogorihm produce a bias.
evaristoc
@evaristoc
Oct 20 2016 21:59
@mesmoiron living in Amsterdam for a while. Ik laat je later meer weten. Ik ga naar bed maar we kunnen later door PM? Tot zo!
Alice Jiang
@becausealice2
Oct 20 2016 23:21
I should move to Amsterdam. For the uh... Culture.
Amelia
@apottr
Oct 20 2016 23:40
yes.. "culture"