Archive

Word embeddings, such as Word2Vec or Glove, are vector representations that capture lexical-semantic properties of words. They constitute a practical way for transferring knowledge between two machine learning models, and they contribute to greatly reducing the learning time required for solving various NLP tasks. There is great practical interest in experimenting with different word embedding models. Neural-based models, due to their flexibility, are a great...

Modern datasets are increasingly collected by teams of agents that are spatially distributed: sensor networks, networks of cameras, and teams of robots. To extract information in a scalable manner from those distributed datasets, we need distributed learning. In the vision of distributed learning, no central node exists; the spatially distributed agents are linked by a sparse communication network and exchange short messages between themselves to directly...

One fundamental challenge of our time is to understand the neuroscience of brain and it’s relation to neurological disease as well as human behaviors. In this talk, I will argue that tools from control systems, dynamics, and network science can have a very important role in both retrieving as well as interpreting data from brain sensors, such as functional magnetic resonance imaging (fMRI) and electroencephalogram...

Automatic trajectory classification has countless applications, ranging from the natural sciences, such as zoology and meteorology, to urban planning and sports analysis, and has generated great interest and investigation. The purpose of this work is to propose and test new methods for trajectory clustering, based on shape, rather than spatial position, as is the case with previous methods. The proposed approach starts by uniformly resampling...

Operation of teams of artificial helpers is one of the hallmarks of technology for the near future, both in hazardous situations and in everyday life. In harsh environments as sea exploration and exploitation, search and rescue operations, and even in most indoor applications as butler robot groups, joint operation is critically supported by cooperative self-localization of each agent, and non-cooperative localization of targets. Networks of agents...

In this talk I will present work presented at EMNLP 2016, about a fast and scalable method for semi-supervised learning of sequence models. The proposed method is based on anchor words and moment matching techniques to retrieve the hidden assignment in a Hidden Markov structure. We can handle feature-based observations. Unlike other semi-supervised methods, we propose a more efficient approach where no additional decoding passes are...

Image denoising is one of the core problems in image processing and some argue that current state-of-the-art general purpose methods may be reaching the maximum possible performance. This talk focuses on two main topics: 1) how can we push such limit; 2) how to leverage the developments in image denoising to tackle other inverse imaging problems. In the former, we propose learning priors that are...

Online news editors ask themselves the same question many times: what is missing in this news article to go online? This is not an easy question to be answered by computational linguistic methods. In this work, we address this important question and characterise the constituents of news article editorial quality. More specifically, we identify 14 aspects related to the content of news articles. Through a...

Scalable Understanding of Multilingual MediA (SUMMA) is an European Horizon 2020 research project which targets to develop a highly scalable platform to automatically monitor public broadcast and web-based news sources, enabling news agencies and journalists to cope with world-scale amounts of information. To this end, a multilingual machine learning stream-processing pipeline is being developed which integrates several technologies such as Audio-to-Speech Recognition (ASR), Machine Translation...

Microsoft started to work on machine learning a few years ago. We incubated a few solutions based in our online platforms and some internal systems. Today we are making them generally available to the public and changing all our products to incorporate intelligent services. In this talk we will look at the past and what influenced the current Microsoft positioning and also look at the...