March 28, 2013 · 6:30 PM
This location is shown only to members
6.30pm Networking and free beer + pizza
7pm Talks start
"Bayesian Neural Networks" by Johanna Fleckner, Project Manager & Data Scientist at Blue-Yonder
A Bayesian Neural Networks approach offers several important features like:
- The conventional approach to network training, based on the minimization of an error function, can be seen as a specific approximation to a full Bayesian treatment.
- Similarly, the technique of regularization arises in a natural way in the Bayesian framework. The corresponding regularization parameters can be treated consistently within the Bayesian setting, without the need for techniques such as cross-validation.
- For regression problems, error bars, or confidence intervals, can be assigned to the predictions generated by a network.
- For classification problems, the tendency of conventional approache s to make overconfident predictions in regions of sparse data can be avoided.
- Bayesian methods provide an objective and principled framework for dealing with the issue of model complexity (for example, how to select the number of hidden units in a feed-forward network), and avoids many of the problems of over-fitting which arise when using maximum likelihood.
Beer and networking break
"Reservoir computing: Adaptive online machine learning and neural networks " by Neri van Otten, Data Scientist at Conversocial
Reservoir computing is a term used to couple together several independently found techniques. The most well known of which are the Echo state network and the Liquid-state machine. These techniques are normally used for offline training but it is possible to also use them for online problems. This talk will demonstrate how
Event finishes by 9.30pm-ish