Hello everyone!
I'm Rachel Reddick, and I'm a data scientist working at Coursera. I'm going to be giving a data science webinar in about a week (on Feb. 7).
If you have questions, I'd love to hear them in advance.
To pique your interest -- I plan to talk primarily about what I've been working on recently (which involves understanding what skills are taught in Coursera's courses). However, if you want to ask questions about other data science topics (such as recommendations I've worked on) or my career path, that's fair game.
I look forward to (virtually) chatting with you!
---
You can find a recording of Rachel's webinar and a PDF of her slides here: Using data science to understand skills learned in Coursera courses (Added by @Laura)
Data Science Webinar -- Ask me things!
Userlevel 1
+1
+1
Hello Rachel!
It seems that this topic touches something that I've been very interested recently, namely knowledge representation. In particular, how was the skill graph developed? How does it look like? Who was working on it? As a self-taught data scientist I have to say that skills related to organising knowledge and building ontologies are hard to build from online resources.
It seems that this topic touches something that I've been very interested recently, namely knowledge representation. In particular, how was the skill graph developed? How does it look like? Who was working on it? As a self-taught data scientist I have to say that skills related to organising knowledge and building ontologies are hard to build from online resources.
Hi Rachel,
I recently started learning about Time Series Analysis. So in your experience, do you think TSA is an important skill that a data scientist should master? And how would you compare TSA with some machine learning algorithms, such as RNN (e.g. LSTM, GRU..) which can also process the sequential data?
Another question is, I've been looking for some volunteering work or part-time jobs to get some experience in the real-life data science projects, do you have any recommendation for where we can find those?
I recently started learning about Time Series Analysis. So in your experience, do you think TSA is an important skill that a data scientist should master? And how would you compare TSA with some machine learning algorithms, such as RNN (e.g. LSTM, GRU..) which can also process the sequential data?
Another question is, I've been looking for some volunteering work or part-time jobs to get some experience in the real-life data science projects, do you have any recommendation for where we can find those?
Hi Rachel,
I would love to know if coursera plans to have data science application based courses for different domains. I am from healthcare domain and I find it difficult to find courses that has applications in this field. ANy recommendations will be much appreciated.
Thanks
I would love to know if coursera plans to have data science application based courses for different domains. I am from healthcare domain and I find it difficult to find courses that has applications in this field. ANy recommendations will be much appreciated.
Thanks
@Nisha that's a great question that I think @Vidisha might be able to answer for you.
I am curious about what ways you might be analyzing the mix of hard and soft skills contribute to a successful learning outcome. Often times educators define the hard skills prerequisites without the soft ones. Or, the hard skill prerequisites defined do not adequately gauge the many hard/soft combinations that have a propensity to lead to a successful learning outcome.
Hi Rachel,
I am pursung a Msc.Computer Sciece in Uni South Florida. Can't you give me some suggestions on how to utilize the school resources to prepare for a job application in data analytic field after graduation? Thanks.
I am pursung a Msc.Computer Sciece in Uni South Florida. Can't you give me some suggestions on how to utilize the school resources to prepare for a job application in data analytic field after graduation? Thanks.
Userlevel 6
+10
@rreddick, here's a question that was sent via email:
I first learned R at Coursera 4 years ago and since I got hooked and migrated from SAS to R as my main tool for data science projects. I would love very much to see all my team adopt and migrate to R as well. However it appears to me that R's popularity is in decline to the benefit of Python.
My question is: How do you see the future of R as a player in the data science community? It pains me to see people lose their enthusiasm for R. It's such a great tool. I'd appreciate any insight or comments.
I first learned R at Coursera 4 years ago and since I got hooked and migrated from SAS to R as my main tool for data science projects. I would love very much to see all my team adopt and migrate to R as well. However it appears to me that R's popularity is in decline to the benefit of Python.
My question is: How do you see the future of R as a player in the data science community? It pains me to see people lose their enthusiasm for R. It's such a great tool. I'd appreciate any insight or comments.
+1
First of all, thank you @rreddick for your excellent webinar!
I have a couple of questions/discussion points for anyone who has worked on this type of projects:
1) Model evaluation
How to measure the effectiveness of such a model? With a simple model such as GBM (I'm assuming it works on BOW on n-grams?), tens of thousands of labels, messy and wildly imbalanced data, the precision and accuracy are probably quite low. A/B testing? But how to conduct them effectively? I also find that relying purely on A/B testing makes iterating on model design much slower than desired.
2) Perverse incentives
When asking those who make courses to provide labels, aren't they incentivized to provide as many as possible, just to increase their courses' reach? I would think that even when preventive measures are taken, some people might still try to "game" the system.
Very curious to see what people think!
(Moved to this thread by @Laura)
I have a couple of questions/discussion points for anyone who has worked on this type of projects:
1) Model evaluation
How to measure the effectiveness of such a model? With a simple model such as GBM (I'm assuming it works on BOW on n-grams?), tens of thousands of labels, messy and wildly imbalanced data, the precision and accuracy are probably quite low. A/B testing? But how to conduct them effectively? I also find that relying purely on A/B testing makes iterating on model design much slower than desired.
2) Perverse incentives
When asking those who make courses to provide labels, aren't they incentivized to provide as many as possible, just to increase their courses' reach? I would think that even when preventive measures are taken, some people might still try to "game" the system.
Very curious to see what people think!
(Moved to this thread by @Laura)
Hi Rachel,
Is there any video record of the webinar?
Thanks!
Is there any video record of the webinar?
Thanks!
Userlevel 6
+10
Hi @Carlos Santoro. You can find the recording here: Using data science to understand skills learned in Coursera courses
(I've now added it to Rachel's post, as well. Thanks!)
(I've now added it to Rachel's post, as well. Thanks!)
Reply
Help
Shortcuts
Bold
Ctrl-B or Cmd-B
Italic
Ctrl-I or Cmd-I
Underline
Ctrl-U or Cmd-U
@Mention
@username
Upload image
Maximum file size: 2MB. Allowed image types: .jpg, .png, .gif.
Embed media
Supported platforms: YouTube, Soundcloud, Deezer, Vimeo, Dailymotion.
Enter your username or e-mail address to receive an e-mail with instructions to reset your password.
Coursera provides universal access to the world’s best education, partnering with top universities and organizations to offer courses online.
© 2018 Coursera Inc. All rights reserved.
