Microsoft Student Partners – Azure Machine Learning Workshop at University of Cambridge

Guest Blog from Microsoft Student Partner Lars Hulstaert

On the 3rd of February, the Cambridge MSP’s organised the first workshop in a series of MSP-led workshops. The goal of these workshops is to provide a short introduction to a topic (e.g. machine learning, IoT, etc.) and provide the attendees with some hand-on experience using Microsoft tools. The Azure framework often provides very accessible tools that can be used to explain and demonstrate new technology – a great example of this is the Azure ML Studio which allows users to drag, drop and connect different components together without any knowledge of coding.

Workshop

A key part of advertising the event, involved setting up a ‘Microsoft Cambridge Events’ page on Facebook, for creating events and uploading pictures. This was shared via multiple Cambridge Computer Science pages and gained a lot of traction fairly quickly. After the event, we also used it to attain feedback and distribute Azure codes for people who had to leave the workshop early.

Approximately 70-80 people attended the workshop, both from the Computer Science department and the Engineering department which is located on the other side of the city.

Overall the feedback was very positive and in order to learn from each other’s experience when organising workshops and events as MSP’s, I decided to do a short write-up on the workshop. The goal is to come up with a set of lessons learned, which will be useful to students and academic who are interested in workshops on Azure Machine Learning..

Prior to this workshop, we had 2 meetups in order to get to know each other and determine the topics of the workshops. Charlie (the Cambridge MSP lead) had split the MSP’s into two workshop groups, allowing the MSP’s to specialise on certain topics and work efficiently within their group.

As I had already finished the Microsoft Professional Program on Data Science and Machine Learning, I had quite some experience with Azure Machine Learning Studio. I prepared the workshop based on a workshop on the following Github resources https://github.com/MSFTimagine/computerscience

I however decided to redo the slide deck (in attachment) and reformat the workshop slightly to have it fit within a 1h30 slot.

The final workshop consisted roughly out of the following:

- a 15 min introduction to the basic principles of Machine Learning and Data Science, and an overview of the applications
we mainly targeted undergrads and students with limited experience in Machine Learning and Data Science. The introduction aimed to get everybody on common ground.
- an introduction to Azure ML Studio by performing an experiment (digit recognition):

The different components that were needed in the experiment were introduced, without connecting them or showing the complete setup to the attendees

- 30 min for the attendees to run the experiment on Azure ML Studio, upload the dataset, drag-and-drop the components and connect them. This was interrupted every now and then to show a part of the solution, so that by the end of the 30 min, most of the attendees had a completely running experiment

- a demonstration of the Azure ML Studio service running in an application. The original workshop integrated the digit recognition model into a web application. To limit time, I just showed how the experiment can be deployed as a web service and integrated into a web application.