Machine Learning from Octave to R Edward Vanden

  • Slides: 13
Download presentation
Machine Learning – from Octave to R Edward Vanden Berghe

Machine Learning – from Octave to R Edward Vanden Berghe

Bio • Edward Vanden Berghe • Ph. D Science, VUB – Marine Biology •

Bio • Edward Vanden Berghe • Ph. D Science, VUB – Marine Biology • Data scientist avant la lettre – Active in data management and data analysis • Joined the Brussels Data Science Community – From the very first meeting, 27 March 2014 • Joined the European Data Innovation Hub – Umbrella over number of activities, including BDSC – Now with Belgian Road Safety Institute

Coached MOOCs • Massive Online Open Course – Abused for any on-line learning platform

Coached MOOCs • Massive Online Open Course – Abused for any on-line learning platform • Many very interesting courses, but massive drop-out rate • Idea: follow MOOC as a group, ‘coached’ by some one-eye • First course: Machine Learning, Andrew Ng

Introduction Welcome Machine Learning

Introduction Welcome Machine Learning

Structure of Andrew Ng’s class • Intro: video • Slides were made available as

Structure of Andrew Ng’s class • Intro: video • Slides were made available as resources • Excercises: – PDF with explanation and help – ‘Skeleton’ Octave scripts, where parts of the solution were left out and had to be added in by the student – Grading on the basis of theoretical questions, and automatically grading the exercises

Why Octave? • • ‘Neat’ language Open Source Very good for linear algebra/matrix algebra

Why Octave? • • ‘Neat’ language Open Source Very good for linear algebra/matrix algebra Open-source version of Mat. Lab

Why R? • R is the best! • Much more popular than Octave or

Why R? • R is the best! • Much more popular than Octave or Mat. Lab with data science • Large number of libraries, also specific for Machine Learning

Structure of a session • Going through slides (not videos) • Discussing possible problems

Structure of a session • Going through slides (not videos) • Discussing possible problems – With theory – With the Exercises • Distributing the R versions of the skeleton scripts • Next session: – Going through solution part of the scripts – Adding some extra stuff: • Solutions based on standard libraries • Applying both library- and own solutions to extra problems

Why this presentation? • Is there interest in the R community to repeat this

Why this presentation? • Is there interest in the R community to repeat this MOOC? • Or, alternatively, turn this into our own webbased machine learning resource? – Needs cleaning up and expanding the skeleton scripts – Needs more extra exercises • Now: Boston Housing, Titanic – Needs expanding to newer tools • Spark. R, H 2 O, Azure…