We formulate the problem of metric learning for k nearest neighbor classification as a large margin structured prediction problem, with a latent variable representing the choice of neighbors and the task loss directly corresponding to classification error. We describe an efficient algorithm for exact loss augmented inference,and a fast gradient descent algorithm for learning in this model. The objective drives the metric to establish neighborhood boundaries that benefit the true class labels for the training points. Our approach, reminiscent of gerrymandering (redrawing of political boundaries to provide advantage to certain parties), is more direct in its handling of optimizing classification accuracy than those previously proposed. In experiments on a variety of data sets our method is shown to achieve excellent results compared to current state of the art in metric learning.

The US Office of Science and Technology today released more information about President Donald Trump's FY 2021 budget and, as it stands, it looks like the White House and Pentagon are finally getting serious about AI and quantum computing technology. The new budget contains billions of dollars earmarked for AI research and development (R&D) spread out over various government agencies. This includes a record $1 billion set aside for non-military AI research. The National Science Foundation has a budget of $830 million – a 70% increase over FY 2020. Other agencies spending hundreds of millions include the Departments of Agriculture, Energy, and Defense, all of which received increased funding in the new budget.

The US federal government is investing heavily in research on quantum computing, and AI is helping to boost the development. The White House is pushing to add an additional billion dollars to fund AI research that would increase AI R&D funding research to nearly $2 billion and quantum computing research to about $860 million over the next two years, according to an account in TechCrunch on Feb. 7. This is in addition to the $625 million investment in National Quantum Information Science Research Centers announced by the Department of Energy's (DoE) Office of Science in January, following from the National quantum Initiative Act, according to an account in MeriTalk. "The purpose of these centers will be to push the current state-of-the-art science and technology toward realizing the full potential of quantum-based applications, from computing, to communication, to sensing," the announcement stated. The centers are expected to work across multiple technical areas of interest, including quantum communication, computing, devices, applications, and foundries.

So says the famed quote from Shakespeare's The Tempest, alleging that we can look to what has already happened as an indication of what will happen next. This idea could be interpreted as being rather bleak; are we doomed to repeat the errors of the past until we correct them? We certainly do need to learn and re-learn life lessons--whether in our work, relationships, finances, health, or other areas--in order to grow as people. Zooming out, the same phenomenon exists on a much bigger scale--that of our collective human history. We like to think we're improving as a species, but haven't yet come close to doing away with the conflicts and injustices that plagued our ancestors.

Law enforcement agencies around the world are enthusiastically adopting the services of Clearview AI, a tech company whose powerful software scrapes several billion open-source images for the purposes of facial recognition. As the company confronts mounting criticism over its disturbing surveillance practices, its CEO, Hoan Ton-That, is rolling out an audacious new defense: he claims that Clearview's practices are protected by the first amendment. Ton-That's upside-down views of civil liberties are, it seems, just as Orwellian as his company's surveillance apparatus. Fortunately he is dead wrong. The constitution does not shield Clearview AI from accountability.

Recent work on deriving $O(\log T)$ anytime regret bounds for stochastic dueling bandit problems has considered mostly Condorcet winners, which do not always exist, and more recently, winners defined by the Copeland set, which do always exist. In this work, we consider a broad notion of winners defined by tournament solutions in social choice theory, which include the Copeland set as a special case but also include several other notions of winners such as the top cycle, uncovered set, and Banks set, and which, like the Copeland set, always exist. We develop a family of UCB-style dueling bandit algorithms for such general tournament solutions, and show $O(\log T)$ anytime regret bounds for them. Experiments confirm the ability of our algorithms to achieve low regret relative to the target winning set of interest. Papers published at the Neural Information Processing Systems Conference.

"There's this folklore mythology around if Amazon launches a business in a certain area, it means that all the other businesses in those areas are not going to be as successful," Jassy said at the Goldman Sachs Technology and Internet Conference in San Francisco yesterday. "I just haven't seen it." There are only two significant industries that Amazon has "disrupted," according to Jassy: retail with Amazon.com, and technology infrastructure with AWS. His remarks come as federal and state regulators are conducting antitrust probes to determine whether Amazon and other technology giants stifle competition and innovation. "In both cases, they were models that were pretty antiquated, and customers weren't so happy with those models, and somebody was going to end up reinventing them," Jassy said.

For some, fears of AI lie in images of robot overlords and self-aware malware -- the stuff of science fiction. Among the many threats we will deal with in the coming years, sentient AI taking over the world isn't one of them. But AI that empowers cybercriminals is a very serious reality, even as some espouse the benefits of AI in cybersecurity. Over the past decade, advances in technology have reduced the time criminals need to modify malware samples or identify vulnerabilities. Those tools have become readily available, leading to an increase in the development and distribution of "regular" threats, like adware, trojans, and ransomware.

Argument Mining is the research area which aims at extracting argument components and predicting argumentative relations (i.e., support and attack) from text. In particular, numerous approaches have been proposed in the literature to predict the relations holding between the arguments, and application-specific annotated resources were built for this purpose. Despite the fact that these resources have been created to experiment on the same task, the definition of a single relation prediction method to be successfully applied to a significant portion of these datasets is an open research problem in Argument Mining. This means that none of the methods proposed in the literature can be easily ported from one resource to another. In this paper, we address this problem by proposing a set of dataset independent strong neural baselines which obtain homogeneous results on all the datasets proposed in the literature for the argumentative relation prediction task. Thus, our baselines can be employed by the Argument Mining community to compare more effectively how well a method performs on the argumentative relation prediction task.

For organizing large text corpora topic modeling provides useful tools. A widely used method is Latent Dirichlet Allocation (LDA), a generative probabilistic model which models single texts in a collection of texts as mixtures of latent topics. The assignments of words to topics rely on initial values such that generally the outcome of LDA is not fully reproducible. In addition, the reassignment via Gibbs Sampling is based on conditional distributions, leading to different results in replicated runs on the same text data. This fact is often neglected in everyday practice. We aim to improve the reliability of LDA results. Therefore, we study the stability of LDA by comparing assignments from replicated runs. We propose to quantify the similarity of two generated topics by a modified Jaccard coefficient. Using such similarities, topics can be clustered. A new pruning algorithm for hierarchical clustering results based on the idea that two LDA runs create pairs of similar topics is proposed. This approach leads to the new measure S-CLOP ({\bf S}imilarity of multiple sets by {\bf C}lustering with {\bf LO}cal {\bf P}runing) for quantifying the stability of LDA models. We discuss some characteristics of this measure and illustrate it with an application to real data consisting of newspaper articles from \textit{USA Today}. Our results show that the measure S-CLOP is useful for assessing the stability of LDA models or any other topic modeling procedure that characterize its topics by word distributions. Based on the newly proposed measure for LDA stability, we propose a method to increase the reliability and hence to improve the reproducibility of empirical findings based on topic modeling. This increase in reliability is obtained by running the LDA several times and taking as prototype the most representative run, that is the LDA run with highest average similarity to all other runs.