Fraser and Sabine Transcript

transcripts/sabine-and-fraser-2014.md

@@ -0,0 +1,53 @@
+Sabine Crevoisier & Fraser Campbell 
+
+###The Open-Source Masters
+
+After realising that everything we wanted to learn was online for free in brilliant open source courses, we spent 6 months in Northern Thailand becoming Data Scientists. At the end, we built a website visualising what we learnt:
+[Data Science Journey](http://datasciencejourney.com/)
+
+###Our Backgrounds
+
+We both originally studied Physics at Cambridge University before becoming an engineer and analytics consultant respectively. However after being introduced to the world of online education and hearing about machine learning at our work we began a journey to become data scientists.
+
+###Goals & Motivations of the Open Source M.S.
+
+Having worked as analysts before, we realised there was so much potential in data, if only we had the full technical skillset and statistical grounding to confidently unlock it. Therefore we began studying evenings and weekends to learn. But after realising the length of time this would take, we decided to commit ourselves fully for 6 months mimicking the structure of a masters course. The decision of open source relative to a conventional masters was based on:
+* 1. Cost
+* 2. Not as pressurised to attain a certificate due to prior education already achieved
+* 3. Outstanding quality of teaching and a focus on latest tools and techniques
+
+###Courses Completed Prior to studying full-time
+* [Computing for Data Analysis](https://www.coursera.org/course/compdata) Coursera course from John Hopkins University - Learn how to program in R and how to use R for effective data analysis.
+* [Data Analysis](https://www.coursera.org/course/dataanalysis) Coursera course from John Hopkins University - Introduction to data analysis in the R statistical language.
+* [Machine Learning](https://class.coursera.org/ml-004) Coursera course based on Andrew Ng's Machine Learning course taught at Stanford University.
+* [Programming Methodologies](http://see.stanford.edu/errors/default.aspx?aspxerrorpath=/see/courseinfo.aspx) Stanford University's introductory course to functional and object oriented programming using Java.
+* [Statistics One](https://class.coursera.org/ml-004) Coursera course based on Andrew Conway's course taught at Princeton University aimed at teaching the fundamental concepts in statistics.
+
+###The Data Science Curriculum / May-November 2014
+* [Programming Abstractions](http://see.stanford.edu/errors/default.aspx?aspxerrorpath=/see/courseinfo.aspx) Stanford University's follow-up course to programming methodologies, teaching algorithms, recursion and software design.
+* [Natural Language Processing](https://class.coursera.org/nlangp-001) Coursera course on the application of computational models to text or speech data, based on Michael Collins's course from Columbia University.
+* [Statistics](https://www.udacity.com/course/viewer#!/c-st095) Udacity courses introducing descriptive statistics, study design, t-tests, p-values and ANOVA.
+* [Stats 110](http://projects.iq.harvard.edu/stat110/home) A comprehensive introduction to statistics starting with probability theory. The course has a Bayesian focus and introduces the tools required for understanding statistical inference.
+* [Databases](https://class.coursera.org/db) Coursera course on Databases, Relational Algebra and the SQL language. It also contains chapters on XML and NoSQL.
+* [Visualisation](http://www.cs171.org/#!index.md) Harvard University course on visualising information. It teaches and makes extensive use of the D3.js Javascript library to build interactive web visualisations while covering the theory behind human visual cognition.
+* [Data Science](http://cs109.org/) Harvard University course covering the broad range of skills required to undertake projects as a Data Scientist and communicate your results. The Python development environment was taught and used extensively.
+* [Probabilistic Systems Analysis and Applied Probability](http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-041-probabilistic-systems-analysis-and-applied-probability-fall-2010/) We completed the final part of this MIT course, in order to practice statistical inference learnt in Stats 110 above.
+
+###Projects
+* Created an interactive website using D3.js, and did some Bayesian statistical analysis of our study pattern:
+[Data Science Journey](http://datasciencejourney.com/)
+* Predicted ad click through rates using Kaggle platform. The GB sized training data enabled us to practise using an [online learner](http://hunch.net/~vw/) algorithm. / [ipython notebook](http://nbviewer.ipython.org/github/fraser-campbell/Machine-Learning-Projects/blob/master/Avazu/Avazu%20Click%20Through%20Rate.ipynb)
+* Bike rental prediction using Washington DC data provided by Kaggle. We used linear and log-linear regression methods as well as random forest algorithms to train our model. / [ipython notebook](http://nbviewer.ipython.org/github/fraser-campbell/Machine-Learning-Projects/blob/master/Bike%20Sharing/Bike%20Sharing.ipynb)
+
+###Some books we read or are still reading
+* [Bayesian Methods for Hackers](https://github.com/CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers) A practical look at MCMC simulations with python libraries
+* Bayesian Data Analysis, Andrew Gelman: Book on applying Bayesian Methods
+* [Interactive Data Visualisation](http://alignedleft.com/tutorials/d3) Using D3.js to build visualisations in Javascript
+* [Data Science with Open Source Tools](http://it-ebooks.info/book/624/) Applications of Data Science in R and Python from a practitioner
+* Design for Information, Isabel Meirelles
+* [The Elements of Statistical Learning](http://statweb.stanford.edu/~tibs/ElemStatLearn/) This is often cited as the go-to reference for a rigorous mathematical explanation of stats and machine learning methods
+* [Programming Abstractions Course Reader](http://web.stanford.edu/class/cs106l/course-reader/full_course_reader.pdf) Textbook following Stanford's C++ Programming course
+
+The most valuable resources we found to practise what we learned, and learn more from practitioners was [Kaggle](https://www.kaggle.com/)
+
+#### The cost of this was 200+ coffees and we will always be grateful to the institutions that made education available to whoever would seize it.