Artificial Intelligence and Machine Learning at ICAS
A glimpse on some of the AI achievements @ ICAS
Artificial intelligence and Machine Learning at ICAS is powered by a triple-sided engine in which each side is decisive and critically influential for and by the other two. We work with the very basic of Machine Learning in basic science research (papers in particle physics, astronomy, cosmology, weekly seminars and atheneum, workshops, etc). We develop AI tools and applications for Society, Governments, International Organisms and private companies. We teach all levels of AI and Machine Learning at the University, at Government and at private and international Institutions. All three legs hold one to the other and are crucial for a robust and solid growth of the Artificial Intelligence and Machine learing side at ICAS.
International Conferences, Workshops, Contests, seminars and atheneum (Ext-links: [1] [2] [3] [3] (winner at [4]) [5])
Bayesian Inference on Argentina's social data
Socio-Economic index at AMBA (click to further explore)
Educational index at Buenos Aires (click to further explore)
Commercial index at Buenos Aires (click to further explore). Here we mean high level of commercial activity regardless of educational level.
We have used Bayesian Inference techniques to learn the insight of the data structure of the Argentina Census. This is mainly done using factor analysis and matrix-factorisation. Using these ersults we were able to construct self-learned indicators that can track the wealth of the society actoss the spatial dimensions of Buenos Aires. We have also constructed, among others, indices that track the education of the people according to their residence, and an indicator that we've dubbed as commercial index; meaning wealth regardless of educational level. One can explore these three interactive maps in the following links:
Intelligent Arxiv (IArxiv.org): a scientists' recommendation App for everyday papers
Scientists from all around the World on Twiiter
IArxiv can send you a mail every night with the day's papers, and still learn while you click in each paper you like
Every day scientists around the World publish many hundreds of papers at Arxiv.org. Usually in each scientist's field there may be like 20-50 papers every night. Each scientist needs to at least go through their titles/abstracts to be in the state-of-the-art of what is going on and what is being published. This is a very time- and effort- consuming task. We ahve therefore invented Intelligent Arxiv (IArxiv.org), which does much of the work for the scientists. It can recognize what is your Topic preferences in each paper, and build a Topic vector for each scientist, and then sort the papers every night according to each scientist preferences. We use Latent Dirichlet Allocation, a Bayesian framework to extract the Topic composition of the corpus of papers, and then the Topic composition of each paper. More details can be found in our paper about it: Intelligent Arxiv: Sort daily papers by learning users topics preference.
We are very proud to have been chosen by the original Arxiv to be their AI-Engine for sorting paper for users, read the Arxiv-blog for more details.
World-wide scientists use IArxiv.org every day
Work done by Ezequiel Alvarez, Cesar Miquel (Easytech), Federico Lamagna (IB) and Manuel Szewc. Upgrade expanded the team to Marcelo Calarrota (FIUBA), Guillermo Recalde (FIUBA) and Sebastian Campoamor (FIUBA).
Intelligent Parliament: an AI App for the National Congress @Argentina
A few videos during the Intelligent Parliament App presentation
The Intelligent Parliament App performs a Natural Language Processing algorithm over all Congress-men and Congress-women speechs to create a Topic Model that contains the main topics of what is dicussed at the National Congress. It then provides a diverse catalogue of tools to analyze and study the Congress speeches evolution and behaviour along the last 20 years Congress Sessions. You can explore and play with the App in the following link, and also watch some interesting videos on AI.
Work done by Ezequiel Alvarez, Daniel de Florian, Ignacio Borsa, Ivan Pedroni, and from the National Congress team German Tarasewicz, Juan Manuel Cheppi, Solanghe Gonzalez, Diego M., Marina Rouco, and Lucia Tenaglia.
Private national and international coachings and courses on AI
Since 2020 we give lectures, courses and targetted trainings on AI to medium and large scale companies in Argentina and around the World. We have a veriety of courses, mostly by Zoom in thesse cases. We have lectures for executives (learn AI without the need of writing one line of code), for developers (in-depth programs), and also for medium-level developers (learn to use specific algorithms with very hands-on examples). Among our most important customers we have coached in AI JPMorgan Vice-Presidents and Executives (America and Europe), JPMorgan developers (worldwide, 5 courses), Sancor Corporation (medium-level and developers), among others.
Zoom lectures go with very hands-on examples as in this slide, in which we learn how to construct a Machine Learning algorithm to predict house prices in California
Work done by Rodrigo Díaz, Martin Makler, Manuel Szewc, Carla Bonifazi, Ignacio Fabre, Agustín Nieto, Luca Palma, Ivan Pedron, Leo Ermann, and others. Coordinated by Daniel de Florian and Ezequiel Alvarez.
ICAS Campus Miguelete, 25 de Mayo y Francia. C.P.: 1650. San Martín, Provincia de Buenos Aires, Argentina.
Phone: +54 (11) 4006-1500.
icas@unsam.edu.ar
