Avoiding algorithm stagnation…A KACE STORi
Here at Kognitio we don’t believe in standing still when it comes to developing new data models and advanced analytics in particular. With this in mind the KACE team made a recent trip up to Lancaster University where they had been invited to run a STORi workshop. This was an excellent opportunity for us to engage with the data scientists of tomorrow and we chose to present some of the complexities of designing parallel algorithms to handle modelling with big data sets.
There was a wide range of attendees from under-grads to post doc researchers so we took a well-known problem: K-means clustering and asked them to think about parallelising it to minimise passes over the data. Particular areas of discussion included cluster initialisation and amalgamation of results from multiple parallel threads.
Kognitio got some great ideas that we’ll carry into our future algorithm development and the attendees enjoyed the exposure to genuine big data problems. Kaylea Hayes, 3rd year STORi PhD student said
I was facilitating a group of STOR-i interns. In my opinion the problem was very interesting and accessible for all backgrounds and previous knowledge. A “live demo” was very interesting as it gave us an insight into how the problem we were thinking about fitted in to “the real world”… I liked how it was very clearly a real problem for Kognitio.
For more details read the full article here