EPFL Extension School — Applied Data Science: Machine Learning
I have just successfully defended my end-of-course project for the EPFL Extension School and earn an EPFL Certificate of Open Studies Diploma and 15 ECTS credits.
My journey started in May, last year. The lockdown was announced and after a few month of hard work (I’m a business continuity manager) I was a permanent work from home. Time for me to look after an online training and immediately my choices focused on something around data and programming.
I had a look on internet and found some online training website (Udemy, Coursera, Udacity, DataCamp) who proposed very good courses for a valuable prices. So I took one from Udemy and did it very fast … wouha … it’s what I wanted :-)
By chance my company organized a few days later an EPFL — Extension School webinar and I registered. Many training were presented and I decided to apply for the Applied Data Science: Machine Learning course even if the expected quantity of hours was announced to 450 (short calculus, as if you want to complete the cursus in 1 year — which was my target is to spent ~50 hours per month — 12 hours per week).
Registration and beginning of the course.
Registration was facilitated by my company and fees covered … good!.
The training is divided in 4 courses, each course contains subjects and units. A unit is a topic you can complete in a window from 15 to 60 minutes with readings, quiz and most of the time Python Notebooks to reproduce. I like the format of max 1 hours units as you can do one at any time (lunch, travelling, …). The units are relatively independent (need to do it in order) but we can make pause between each unit (I made a few breaks for vacation, golfs or simply hard work).
The first 2 courses covers Python as a programming language, statistics and probability and a large range of data manipulation technics which is big part of Machine Learning discipline (EDA).
Each course is ended by a project (~40/60 hours duration). Along the units we used to work with famous ML dataset as the Titanic’s, Diabetes, House Price and many other that we can find on platform such as Kaggle.
Deep dive in Machine Learning
The 3rd course made a wide introduction in Machine Learning technics and started we some basic concept about linear regression. Times series is covered and all elements as cost functions or model tuning as well.
The 4th module go deeper in machine learning with an introduction of the Deep Learning and usage of Tensorflow, some advanced technique with Sklearn library and units covering chatbot or image processing.
On very positive point, the cursus isn’t focus on finance, medical or other topics but gives material to easily transpose the learning material to our own domain.
During the training, I had access to a weekly 1:1 (30’) where I was able to ask questions, have a general talk and speak about courses. I took a few all along the 12 months.
Capstone project & end of the cursus
At the end of the training, I had to realize a end-to-end project. The topics is free of choice but must be accepted by the learning process. I choose to predict probability of flows (in/out) in portfolio by transforming a time series in a supervised learning dataset and apply classification modeling. The project must cover all material teaches during the course such as data collection, EDA, visualization, modeling and evaluation. I spent 3 months on my project and I graduated last week.
Spending time on the units and training material is not enough. I learned a lot by reading article on Medium and other publishing platform.
I bought 3 books related to machine learning process and 1 specially on Time Series to cover my Capstone Project.
Finally, as it was a online training I tried to get the rythme by spending time each week on my course and some Notebooks programming. For example, I made 2 competitions on Kaggle.