Personal tools
You are here: Home Events ANC Workshop Konstantinos Georgatzis/Scott Lowe Chair Guido Sanguinetti

ANC Workshop Konstantinos Georgatzis/Scott Lowe Chair Guido Sanguinetti

— filed under:

What
  • ANC Workshop Talk
When Mar 24, 2015
from 11:00 AM to 12:00 PM
Where Room IF 4.31/4.33
Add event to calendar vCal
iCal

Details re Konstantinos Georgatzis' talk

TITLE:   "Discriminative Switching Linear Dynamical Systems
              applied to Physiological Condition Monitoring".

ABSTRACT:   I will present a Discriminative Switching Linear Dynamical System (DSLDS) applied to
patient monitoring in Intensive Care Units (ICUs). This approach is based on identifying
the state-of-health of a patient given their observed vital signs using a discriminative
classifier, and then inferring their underlying physiological values conditioned on
this status. The work builds on the Factorial Switching Linear Dynamical System (FSLDS) which
has been previously used in a similar setting.  It is demonstrated on two real-world datasets
that the DSLDS is able to outperform the FSLDS in most cases of interest, and that an alpha-
mixture of the two models achieves higher performance than either of the two models separately.

 

Details re Scott Lowe's talk

TITLE:  Classifying Plankton Species with Computer Vision and Deep Learning

ABSTRACT:  I will discuss the outcome of a recent Kaggle competition "Predict ocean health, one plankton at a time", in which I competed in alongside a team of students from the Informatics department.

The goal of the competition was use machine learning techniques to classify plankton species based on images collected from automated underwater microscopy. We placed 57th out of 1049 teams.

As part of our solution we used convolutional neural networks, hierarchical modelling, and feature-based computer vision methods. I will go over our strategy, both what worked and what didn't, and what we should've done differently to win the $100,000 prize.

For more information on the competition, see:
http://www.kaggle.com/c/datasciencebowl