Recently I gave a talk on the present and future of machine / deep learning (ML / DL)in the enterprise. In an enterprise setting, the questions and topics are more applied than a research conference — for example how do I get my team started in ML and how best to integrate ML with existing systems we run. Afterwards, in the panel discussion the topic of Java and machine learning came up.
Java is noticeably absent in the machine learning space. Virtually no ML frameworks are built in Java (there is DL4J but I genuinely don’t know anyone who uses…
I’ve been wanting to try out the Movidius Neural Compute Stick (NCS) for a while now and decided to combine it with a Raspberry Pi 3 — the idea being of course to move the setup to a mobile-capable chassis (wheeled or winged!) in the near future.
They are selling like hot cakes here in the UK — either that or the manufacturing batches are small. RS Online got their latest batch in last week and they are sold out already -they’ve sold out twice in the last couple of months..
Hardware-wise, £120+VAT gets you a NCS, a Raspberry Pi…
Yesterday I attended the BMVA technical meeting on Human Activity Recognition and Monitoring. I have a fairly atypical motivation to attend events like these as I’m not a computer vision researcher, but am using Machine Learning (currently comparing Deep Learning classifiers to other ML classifier types like Gradient Boosted Machines). It’s interesting to attend ML events that are not directly related to my work — they often engender new ideas and perspectives to help me break out of whatever local minimum I currently find myself in!
At the risk of making a gross simplification, Deep Learning bifurcates into:
I’ve been in the market for a winter bike for a while now. My main bike is a B’TWIN Ultra 720 AF which I’m really happy with but it does have a few drawbacks for winter use — it doesn’t have mudguard eyelets, maxes out at 28 mm tyre width and most importantly of all, doesn’t have disc brakes which are all-important when it’s wet underfoot.
I waited patiently for B’TWIN to release a disc braked bike, but when they did it wasn’t really what I wanted — no mudguard eyelets and a SRAM groupset, instead of Shimano (and I…
14 Jul 2015
This blog post is a more informal discourse on a formal exercise I’m currently engaged in — the background literature review for my PhD. As such, it is very link-rich — the journey I am documenting here is intrinsically made up from the work of others. This post may also be interpreted as an attempt to convince my supervisors Prof. Omer Rana and Prof Ronan Reilly that I am actually getting stuff done instead of reading papers and debugging Torch7 / Theano code. But this is not the case :)
17 Oct 2016
The addition of working memory to artificial neural networks (ANNs) is an obvious upgrade when we compare ANNs to the Von Neumann CPU architecture, and one that came to the fore in the RAM (Reasoning, Attention, Memory) workshop at NIPS last year (which was packed to the rafters). Obvious and building however are two different things..
The recent paper from Google Deepmind on Differentiable Neural Computers or DNCs represents another significant step in the journey to add working memory to ANNs, so it’s worth taking a more in-depth look at it.
In homage to the newly-minted nobel…
Writing is good for the soul