5 May 2014
DEEP LEARNING WORKSHOP
DATE: 6th May 2014
LOCATION: Informatics Forum
With the recent purchase of DeepMind Technologies by Google, deep learning methods are getting increased attention in many spheres. At the same time a number of members of the Scottish Computer Science community have been developing or working with deep learning methods for some time.
Deep Learning is a transformative modelling technology in machine learning. Historically, it was well understood that the process of machine learning involved deciding on a representation, and then applying machine learning tools. The process of deciding a representation was generally manual, involving feature construction, variable selection, data reduction, variable rescaling and transformation, and feature selection. The reality of machine learning was that the real performance of a system was more determined by these manual representation steps than by the machine learning methods that were subsequently applied.
Hence, the issue of automated learning of representations was seen as a key process for improving the power and capability of machine learning methods. Deep Learning methods have come to the fore in this process by providing automated unsupervised methods for representation learning. These unsupervised methods work via iterated layerwise refinement of the previous representations to capture multiscale features that are useful across many tasks. This process of using further and further levels of abstraction, while controlling for abstractions that explain the data is what makes deep learning methods so versatile.
On the 6 May 2014 we will be running a workshop on Deep Learning to bring together those who are working in Machine Learning, Statistics, and the wider community, who are working on, using or interested in Deep Machine Learning Methods. These include methods such as Deep Belief Networks, Deep Boltzmann Machines, Deep Convolutional Networks, Deep Autoencoders or any other related methods, including other methods of automated representation learning. This workshop will be a benefit to those currently utilising or developing such methods, those wanting to know what can be done and understand some of the practical issues hidden behind the hype, or those who are just interested in finding out more. It will also be of interest to those in early stage businesses who wish to harness this exciting technology.
The workshop will be in the School of Informatics, University of Edinburgh. The outline of the event is:
This workshop is funded by the Scottish Informatics and Computer Science Alliance (SICSA), the Institute for Adaptive and Neural Computation (ANC) and the Institute for Language, Communication and Cognition (ILCC) of the School of Informatics, University of Edinburgh.
The workshop registration and workshop lunch is free of charge to registered attendees.
Participants will need to make their own travel arrangements. By train the nearest station is Edinburgh Waverley, which is less than a 15 minute walk from the forum. See National Rail Enquiries for train information. For those coming from further afield, information about travel to and from Edinburgh Airport is available. A taxi to the city centre from the airport costs about 20GBP to 22GBP one way. There is also an excellent express bus from the airport called Airlink that terminates in the city centre and costs 7GBP for a return journey. The journey to the airport requires approximately 30 minutes from the city centre of actual journey time (add a little more during the rush hours).
If you wish to contribute a talk or poster to this workshop, then please send a title and abstract to amos+deep@@inf.ed.ac.uk, replacing the dual @ sign. Please send contributions before 1 April. We will then be in contact about your submission. It is likely that not all submissions will be able to be included.
The workshop is now full. If you are planning not to attend after all, then please deregister to allow others to attend.