Reminder: talk at 12:30 - 3D Lung Cancer Segmentation and Representation Learning for Few-Shot Classification
by Announcements of talks＠IDSIA
The talk will take place in room 204 at 12:30 and will last about 30
minutes. Those interested to join us for lunch later will be welcome, but
there will not be free pizzas this time. :)
*Title:
3D Lung Cancer Segmentation and Representation Learning for Few-Shot
Classification
*Abstract:
The recent advances in Deep Learning made many tasks in Computer Vision
much easier to tackle. However, working with a small amount of data, and
highly imbalanced real-world datasets can still be very challenging. In
this talk, I will present two of my recent projects, where modelling and
training occur under those circumstances. Firstly, I will introduce a novel
3D UNet-like model for fast volumetric segmentation of lung cancer nodules
in Computed Tomography (CT) imagery. This model highly relied on kernel
factorisation and other architectural improvements to reduce the number of
parameters and computational load, allowing its successful use in
production. Secondly, I will discuss the use of representation learning or
similarity metric learning for few-shot classification tasks, and more
specifically its use in a competition at NeurIPS 2019 and Kaggle. This
competition aimed to detect the effects of over 1000 different genetic
treatments to 4 types of human cells, and published a dataset composed of
6-channel fluorescent microscopy images with only a handful of samples per
target class.
*Speaker:
Henrique Mendonça got a MSc at the University of Zurich in 2015 and has 10+
years experience in designing and developing systems in areas from
real-time embedded systems to high performance distributed applications,
computer vision and machine learning.
4 years, 10 months
PIDSIA Seminar by Henrique Mendonça, Friday 14 Feb 12:30
by Announcements of talks＠IDSIA
*Title:
3D Lung Cancer Segmentation and Representation Learning for Few-Shot
Classification
*Abstract:
The recent advances in Deep Learning made many tasks in Computer Vision
much easier to tackle. However, working with a small amount of data, and
highly imbalanced real-world datasets can still be very challenging. In
this talk, I will present two of my recent projects, where modelling and
training occur under those circumstances. Firstly, I will introduce a novel
3D UNet-like model for fast volumetric segmentation of lung cancer nodules
in Computed Tomography (CT) imagery. This model highly relied on kernel
factorisation and other architectural improvements to reduce the number of
parameters and computational load, allowing its successful use in
production. Secondly, I will discuss the use of representation learning or
similarity metric learning for few-shot classification tasks, and more
specifically its use in a competition at NeurIPS 2019 and Kaggle. This
competition aimed to detect the effects of over 1000 different genetic
treatments to 4 types of human cells, and published a dataset composed of
6-channel fluorescent microscopy images with only a handful of samples per
target class.
*Speaker:
Henrique Mendonça got a MSc at the University of Zurich in 2015 and has 10+
years experience in designing and developing systems in areas from
real-time embedded systems to high performance distributed applications,
computer vision and machine learning.
*Date:
Friday, 14th of February 2020, 12:30-13:30
*Location:
Manno, Galleria 1, 2nd floor, room TBA
*Doodle registration:
Pizza (or alternative food) and drinks will be offered at the end of the
talk. If you plan to attend the lunch, please register in a timely fashion
(meaning at the latest at 9AM of the same day of the talk) at the following
link so that we will have no shortage of food.
https://doodle.com/poll/cdqq4ygs56qzrkpb
4 years, 10 months
PIDSIA Seminar by Lilith Mattei (13th of February)
by Announcements of talks＠IDSIA
*Title:
Introducing probabilistic sentential decision diagrams and their credal extension
*Abstract:
Probabilistic sentential decision diagrams are a class of probabilistic graphical models natively embedding logical constraints within a “deep” layered structure with statistical parameters. They thence induce a joint probability distribution over the involved Boolean variables that sharply assigns probability zero to states inconsistent with the logical constraints. In this presentation, I will first introduce and motivate such probabilistic circuits. I will then present a set-valued generalisation of the probabilistic quantification in these models, that allows to replace the sharp specification of the local probabilities with linear constraints over them, In doing so, a (convex) set of joint probability mass functions, all consistent with the assigned logical constraints, is induced.
*Speaker:
Lilith Mattei graduated in pure mathematics at EPFL. After a period of teaching in high school, in 2018 she started working at IDSIA as a research assistant. She is currently a PhD student under the supervision of Alessandro Antonucci and Alessandro Facchini. Her research is in PGMs and logic.
*Date:
Thursday, 13th of February 2020, 12:00-13:00
*Location:
Manno, Galleria 1, 2nd floor, room G1-201
*Doodle registration:
Pizza (or alternative food) and drinks will be offered at the end of the talk. If you plan to attend the lunch, please register in a timely fashion (meaning at the latest at 9AM of the same day of the talk) at the following link so that we will have no shortage of food.
https://doodle.com/poll/zkuyct5f9nq57y6i
If by any chance, after registering for a Pizza at the link above, you know you will not be able to attend the lunch, please cancel your registration asap.
4 years, 10 months