This paper describes a purely data-driven solution to a class of sequential decision-making problems with a large number of concurrent online decisions, with applications to computing systems and operations research. We assume that while the micro-level behaviour of the system can be broadly captured by analytical expressions or simulation, the macro-level or emergent behaviour is complicated by non-linearity, constraints, and stochasticity. If we represent the set of concurrent decisions to be computed as a vector, each element of the vector is assumed to be a continuous variable, and the number of such elements is arbitrarily large and variable from one problem instance to another. We first formulate the decision-making problem as a canonical reinforcement learning (RL) problem, which can be solved using purely data-driven techniques. We modify a standard approach known as advantage actor critic (A2C) to ensure its suitability to the problem at hand, and compare its performance to that of baseline approaches on the specific instance of a multi-product inventory management task. The key modifications include a parallelised formulation of the decision-making task, and a training procedure that explicitly recognises the quantitative relationship between different decisions. We also present experimental results probing the learned policies, and their robustness to variations in the data.

We present a method for characterizing the usage patterns of clinical concepts among different document types, in order to capture semantic differences beyond the lexical level. By training concept embeddings on clinical documents of different types and measuring the differences in their nearest neighborhood structures, we are able to measure divergences in concept usage while correcting for noise in embedding learning. Experiments on the MIMIC-III corpus demonstrate that our approach captures clinically-relevant differences in concept usage and provides an intuitive way to explore semantic characteristics of clinical document collections. 1 Introduction Sublanguage analysis has played a pivotal role in natural language processing of health data, from highlighting the clear linguistic differences between biomedical literature and clinical text (Friedman et al., 2002) to supporting adaptation to multiple languages (Laippala et al., 2009). Recent studies of clinical sublanguage have extended sublanguage study to the document type level, in order to improve our understanding of the syntactic and lexical differences between highly distinct document types used in modern EHR systems (Feldman et al., 2016; Gr on et al., 2019). However, one key axis of sublanguage characterization that has not yet been explored is how domain-specific clinical concepts differ in their usage patterns among different document types.

Parametric models, and particularly neural networks, require weight initialization as a starting point for gradient-based optimization. In most current practices, this is accomplished by using some form of random initialization. Instead, recent work shows that a specific initial parameter set can be learned from a population of tasks, i.e., dataset and target variable for supervised learning tasks. Using this initial parameter set leads to faster convergence for new tasks (model-agnostic meta-learning). Currently, methods for learning model initializations are limited to a population of tasks sharing the same schema, i.e., the same number, order, type and semantics of predictor and target variables. In this paper, we address the problem of meta-learning parameter initialization across tasks with different schemas, i.e., if the number of predictors varies across tasks, while they still share some variables. We propose Chameleon, a model that learns to align different predictor schemas to a common representation. We use permutations and masks of the predictors of the training tasks at hand. In experiments on real-life data sets, we show that Chameleon successfully can learn parameter initializations across tasks with different schemas providing a 26% lift on accuracy on average over random initialization and of 5% over a state-of-the-art method for fixed-schema learning model initializations. To the best of our knowledge, our paper is the first work on the problem of learning model initialization across tasks with different schemas.

Neural network-based systems can now learn to locate the referents of words and phrases in images, answer questions about visual scenes, and execute symbolic instructions as first-person actors in partially-observable worlds. To achieve this so-called grounded language learning, models must overcome challenges that infants face when learning their first words. While it is notable that models with no meaningful prior knowledge overcome these obstacles, researchers currently lack a clear understanding of how they do so, a problem that we attempt to address in this paper. For maximum control and generality, we focus on a simple neural network-based language learning agent, trained via policy-gradient methods, which can interpret single-word instructions in a simulated 3D world. Whilst the goal is not to explicitly model infant word learning, we take inspiration from experimental paradigms in developmental psychology and apply some of these to the artificial agent, exploring the conditions under which established human biases and learning effects emerge. We further propose a novel method for visualising semantic representations in the agent.

Inc. defended the privacy features of its Alexa digital assistant -- and introduced some new tools to reassure users – following months of debate about the practices of the technology giant and its largest competitors. The company plans to roll out a feature that lets users of the Alexa voice-based assistant automatically delete their verbal recordings regularly, on a rolling three-month and 18-month basis. Previously, Alexa users had to manually delete their stored voice recordings on a companion website. "We care about this," Dave Limp, the leader of Amazon's devices and services business, said of privacy during a press event at the company's headquarters in Seattle. "Privacy is absolutely foundational to everything that we do in and around Alexa."

When Google co-founders Larry Page and Sergey Brin were headed to Nevada's Burning Man festival in August of 1998, they wanted users and employees to know they wouldn't be at the search engine's helm for a while. The Ph.D. students at Stanford University decided to replace the second'O' in Google's homepage logo with a stick figure resembling the festival's logo. "It was a little bit of a joke," Jessica Yu, the Google Doodle team lead, tells TIME. "It has definitely evolved a lot since then." What began as a joke became Google Doodles that celebrate and honor holidays, people and issues worldwide, now an important venture for the tech giant.

The ERL Smart Cities Robotics Challenge 2019 takes place from 17-21st September in Milton Keynes, United Kingdom. Funded by the European Commission under the SciRoc Horizon 2020 project, this new challenge of the European Robotics League (ERL) focuses on the role of robots in smart cities. The competition includes five different episodes or scenarios under three categories: human-robot interaction & mobility, emergency and manipulation. On the second day of the competition teams kept working on improving their scores to secure a place in the finals. Manipulation The SciRoc episodes under this category require robots to achieve manipulation tasks, applying some of the task benchmarks (TBMs) of the ERL Professional and the ERL Consumer Service Robots leagues.

In International Conference on Computer Aided Verification (2015), Springer 449–465.

Proteins are a class of large molecules that are involved in the vast majority of biological functions, from cell replication to photosynthesis to cognition. The chemical structure of proteins is very systematic5--they consist of a chain of atoms known as the backbone, which consists of three-atom (nitrogen-carbon-carbon) repeats known as residues, each of which features a sidechain of atoms emanating from the first carbon. In general, there are 20 different options for sidechains, and a residue with a particular type of sidechain is known as an amino acid (so there are also 20 different amino acid types). For billions of years, the process of evolution has optimized the sequence of amino acids that make up naturally occurring proteins to suit the needs of the organisms that make them. So we ask: Can we use computation to design non-naturally occurring proteins that suit our biomedical and industrial needs? This question is a combinatorial optimization problem, because the output of a protein design computation is a sequence of amino acids. Due to the vast diversity of naturally occurring proteins, it is possible--and very useful--to begin a protein design computation with a naturally occurring protein and then to modify it to achieve the desired function. In this article, we focus on protein design algorithms that perform this optimization using detailed modeling of the 3D structure of the protein.5,8 Thus, they will begin with a starting structure, a 3D structure of a (typically naturally occurring) protein we wish to modify. To illustrate this concept, imagine we wish to perform a simple example modification to a protein to make it more stable, so it can still function at higher temperatures.

There is a growing consensus among policymakers that bringing high-quality evidence to bear on public policy decisions is essential to supporting the effective and efficient government their constituencies want and need. At the U.S. federal level, this view is reflected in a recent Congressional report by the Commission on Evidence-Based Policymaking, which recommends creating a data infrastructure that enables "a future in which rigorous evidence is created efficiently, as a routine part of government operations, and used to construct effective public policy."4 This article describes a new approach to data infrastructure for fact-based policy, developed through a partnership between our interdisciplinary organization Research Improving People's Livesa and the State of Rhode Island.13 Together, we constructed RI 360, an anonymized database that integrates administrative records from siloed databases across nearly every Rhode Island state agency. The comprehensive scope of RI 360 has enabled new insights across a wide range of policy areas, and supports ongoing research into improving policies to alleviate poverty and increase economic opportunity for all Rhode Island residents (see the sidebar "Policy Areas in which RI 360 Has Contributed Insights").