The UdeM-McGill-MITACS machine learning seminar series is continuing its
fall schedule.* Friday*'s seminar:
*Adaptive stochastic search: tuning Gaussians' covariances*
by Rémi Bardenet and Djalel Benbouzid
University Paris-Sud XI
*Location*: Pavillon André-Aisenstadt (UdeM), room AA-3195
*Time*: *Friday, December 3, 10:30 (!)*
*Abstract*: Stochastic search algorithms are ubiquitous in optimization and
statistics: they are at the core of some of the most efficient blackbox
optimization …
[View More]techniques (e.g., Evolutionary Strategies), and of great
practical use in Bayesian inference (e.g., in Metropolis-Hastings
algorithms). In this talk I will concentrate on stochastic search techniques
with Gaussian proposals that learn from their exploration to tune their
covariance. After a quick overview of the field and methods, I will present
1) an evolutionary optimization algorithm with mixture proposals and its
application to a Bayesian optimization problem, and 2) a novel
attempt-in-progress at defining nonlinear adaptive Gaussian proposals. The
first item deals with Gaussian Process based surrogate optimization: when
the function to optimize is costly to evaluate (e.g., hyperparameter
optimization in Machine Learning), one often relies on a surrogate model
that is learnt on the fly, but optimizing this auxiliary surrogate can be a
difficult task itself. The second problem is about mixing good properties of
adaptive MCMC with reproducing kernel Hilbert spaces to exploit nonlinearity
in the data.
[View Less]
The UdeM-McGill-MITACS machine learning seminar series is continuing this
Friday:
*The use of visual information during face recognition and reading*
by Frédéric Gosselin
Département de psychologie
Université de Montréal
*Location*: Pavillon André-Aisenstadt (UdeM), room AA-3195
*Time*: Friday, Nov 26, *14:30*
*Abstract*: I will present a number of studies that were carried out in my
laboratory using Bubbles, a technique that I pioneered almost 10 years ago
(Gosselin & Schyns, 2001). …
[View More]Bubbles is a generic psychophysical technique
designed to reveal the effective information in any recognition task. I will
talk about Bubbles experiments on the use of information to discriminate the
six basic facial expressions (happiness, fear, sadness, surprise, disgust,
and anger) in normal humans (Smith et al., 2005) and in SM, a woman with a
brain lesion encompassing all nuclei of the amygdala bilaterally, which has
difficulty discriminating fearful from happy faces (Adolphs et al., 2005).
For example, I’ve shown that SM’s difficulty co-occurs with a failure to
use high-spatial frequencies (90-22.5 cycles per face) within the eye region
of the face stimuli like normal controls. Relatedly, I've discovered that
unilateral amygdala lesions induce ipsilateral biases in a facial expression
discrimination task but not in a facial gender discrimination task
(Gosselin, et al., in press). I'll briefly discuss other face recognition
tasks that were examined using Bubbles include gender (e.g., Dupuis-Roy et
al., 2009), identity (Willenbockel et al., 2010), and age. Furthermore,
I’ll talk about the application of Bubbles to reveal which areas of letters
are efficient for the identification of letters (Fiset et al., 2008). I’ve
provided the first empirical evidence that line terminations are the most
important features for letter identification. Finally, I’ll present the
space-time use of letter information while reading using the Bubbles
technique (Blais et al., 2009).
[View Less]