Wednesday, October 07, 2015

Tonight: Paris Machine Learning #2 Season 3: Extreme Multi-class, Infosec, RTB and more.

Video taken by the SNIPS folks: 
14:20 Raphael Puget 
42:30 Imad Soldani
1:09:09 Joaquin 
1:33:17 Amine El Helou
1:43:45 Mouhidine Seiv

Video of the streaming (sound is not optimal, same video as above but from a different angle)

Le meetup will be taking place at SNIPS and will be sponsored by Mathworks. Un grand merci à eux ! We will start at 19h15 Paris time.
Right now the program is as follows:

Franck Bardol, 
Igor Carron
Raphael Puget, LIP6, starts a 14 mins 31 s

Extreme multi-class classification with large number of categories / Classification multi-class dans un très grand nombre de catégories.

Abstract: Extreme multi-class classification concerns classification problems with very large number of classes, up to several millions. Such problems have now become quite frequent in many practical applications. Until recently, most classification methods had inference complexity at least linear in the number of classes. Several directions have been recently explored for limiting this complexity, but the challenge of learning an optimal compromise between inference complexity and classification accuracy is still largely open. We propose here a novel ensemble learning approach, where classifiers are dynamically chosen among a pre-trained set of classifiers and are iteratively combined in order to achieve an efficient trade-off between inference complexity and classification accuracy. The proposed model uses statistical bounds to discard during the inference process irrelevant classes and to choose the most informative classifier with respect to the information gathered during the previous steps. Experiments on real datasets of recent challenges show that the proposed approach is able to achieve a very high classification accuracy in comparison to baselines and recent proposed approaches for similar inference time complexity"

R , information security , large protocol inspection and state machine analysis, Imad Soltani,

The PRISMA package ( is capable of loading and processing huge text corpora processed with the sally toolbox ( sally acts as a very fast preprocessor
which splits the text files into tokens or n-grams. Today , with the deployement of specially honeypots , we can work with PRISMA to implement a protocol inspection and state analysis method . And in combination with sally toolbox , we can provide a more deeper analysis

"RTB à la Quant", JFT

Ce talk presente un approche "quant" pour le RTB. Le but c'est de montrer que c'est possible faire du "pricing" et "couverture de strategies" via des techniques type stochastic dynamic programming. On montre comment, sous certaines hypothèses, on peut garantir des performances a l'avance ainsi que savoir combien il doit couter l'algorithme. Ce travail presente une entrée des techniques puissantes de maths financières en RTB ainsi que signaler les points où l'on a besoin d'une formulation type reinforcement learning.

Mouhidine Seiv,

"I would like to present, a startup that I launched during my gap year in 2013-2014 with a deep belief that there are patterns to be discovered and used to improve Employment and Education. At Riminder we believe that Artificial Intelligence is the key to achieve that goal.

Today we are proud to have 200+ companies using Riminder​, a robot which analyzes the recruiter’s habits and the job seekers' CV to enable faster and better quality recruitment.

On 1st September 2015 we released our own artificially intelligent personal digital assistant : Jarvis (Just A Rather Very Intelligent System).
This new Question Answering machine system answers two of our main aspirations :
- Enable the job seekers a more ubiquitous access to information.
- Design and build software that is smart and fun !

The team is based in Paris and supported by some of the most iconic professors and technology entrepreneurs from Ecole Centrale Paris and Ecole Normale Supérieure.

Beta of Jarvis :
LeMonde talking about Riminder :
Amine El HelouMathworks, Upcoming webinar Machine Learning for Sensor Data Analytics, 5/11/2015

Join the CompressiveSensing subreddit or the Google+ Community or the Facebook page and post there !
Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.

No comments: