UPDATE (March 29th) -
Thank you to everyone that expressed interest in attending the Machine Learning Group. It's great to see this much interest and enthusiasm for Sebastian's proposal!
Good news - Sebastian has capacity to run 2 separate Machine Learning Study groups in tandem over the coming months.These groups will consist of members at Expert/Intermediate level - based on answers to the expression of interest RSVP questions (and as such this activity is closed to new RSVPs).
This most experienced respondents to "Expression of Interest" have been invited to attend the kick off meeting in the first instance.
Having twenty in a study group would be unmanageable once reading begins. So those attending on the night will be split into the two smaller study groups for the remaining 7 sessions (we will divide the groups so they have 8-11 participants each)
At this point in time we are unable to cater to beginners. We are looking for ways to do this in the future. So, we are retaining the list of everyone who expressed interest. They would be given first opportunity if a beginner group forms. If anyone not attending would like to lead a study group, please ask them to email Louise or Eugene.
Over the past few weeks, several other members followed up with me to say they wish they'd had a chance to RSVP before the activity closed. Clearly there is a lot of interest in this concept and/or topic. If it works well and there is demand - it can occur again.
UPDATE- We will be taking up to 40 expressions of interest before closing the activity. Due to popular response, Sebastian is considering running two separate groups on different dates - an expert group and a less experienced group. Each group would have a maximum of 11 participants. The expert group will kick off on Tuesday April 9th. *********************
Inviting Sydney suRf members to participate in an 8 session
Machine Learning study group, starting on Tuesday April 9th at the Sydney CBD
Meetings will probably occur every 2-4 weeks. Frequency will be discussed at this meeting. If participants have capacity, we may even be able to meet weekly.
Ideally participants will have :
- good familiarity with R, including R programming experience
- some understanding of Bayesian and frequentist statistics
The group will meet 8 times to discuss some sections of “Machine Learning for Hackers”, by Drew Conway and John Myles White. http://shop.oreilly.com/product/0636920018483.do
There is no charge for this study group, however places are limited (12 places maximum). The tentative program appears at the bottom of the page.
When the group meets, we will briefly review the main concepts and attempt to solve any queries. Then we will share other resources, such as statistical methods or software that could be used in addition to those in the book. A collaborative environment will be encouraged.
By participating in this study group, you will learn some of the more popular machine learning techniques and their implementation in R.
A good motivational video is available here: http://www.infoq.com/presentations/Machine-Learning , and here is an introductory video by the authors http://oreillynet.com/pub/e/2353
To express interest in attending, please RSVP by Thursday March 29th. When you RSVP,we will ask you some questions about your availability and your experience with R. This will enable our study group leader (Sebastian) to make final decisions about the program.
Please ensure you answer all the questions when you RSVP.
***** Tentative Programme *****
Meeting 1 – April 9th 2013 - Informal discussion to get the group started.
This includes reviewing the programme and deciding on a venue, depending on the size of the group and the interests of the attendees.
Meeting 2 - Chapters 1 and 2 in the book: R for machine learning and Data exploration
This includes some graphics in ggplot2 and other useful tool in R
Meeting 3 – Chapters 3 and 4: Classification: Spam filtering, and Ranking: Priority inbox.
These chapters use an example of Naive Bayes applied to email data.
Meeting 4 – Chapter 5 and 6: Regression: predicting pageviews, and Regularization: Text regressions.
This chapter is a quick overview of linear regression and its use in predicting website views. An additional discussion can include poisson and logistic regression.
Meeting 5 – Chapters 7 and 8: Optimization: Breaking codes, and PCA: Building a market index
An additional topic can be single variable decomposition (svd), since PCA is a subcase of SVD.
Meeting 6 – Chapters 9: MDS: Visually Exploring US Senator Similarity
Although the material if not very complicated, this chapter covers extensive data munging. Therefore some discussion on other data manipulation techniques can be relevant at this stage.
Meeting 7 – Chapter 10: kNN: Recommendation systems
The algorithm used in this chapter is not very complicated. But it has many applications and pitfalls, which open room for more technical discussion.
Meeting 8 – Chapters 11 and 12: Analyzing Social Graphs, and Model Comparison
This includes Support vector machine and web scraping, which might be of special interest. If necessary, this meeting can be extended to two sessions.