ABSTRACT
Intuitively, experience playing against one mixture of opponents in a given domain should be relevant for a different mixture in the same domain. If the mixture changes, ideally we would not have to train from scratch, but rather could transfer what we have learned to construct a policy to play against the new mixture. We propose a transfer learning method, Q-Mixing, that starts by learning Q-values against each pure-strategy opponent. Then a Q-value for any distribution of opponent strategies is approximated by appropriately averaging the separately learned Q-values. From these components, we construct policies against all opponent mixtures without any further training. We empirically validate Q-Mixing in two environments: a simple grid-world soccer environment, and a social dilemma game. Our experiments find that Q-Mixing can successfully transfer knowledge across any mixture of opponents. Next, we consider the use of observations during play to update the believed distribution of opponents. We introduce an opponent policy classifier-trained reusing Q-learning data-and use the classifier results to refine the mixing of Q-values. Q-Mixing augmented with the opponent policy classifier performs better, with higher variance, than training directly against a mixed-strategy opponent.
ABSTRACT
Early during the COVID-19 pandemic, the Centers for Disease Control and Prevention (CDC) leveraged an existing surveillance system infrastructure to monitor COVID-19 cases and deaths in the United States. Given the time needed to report individual-level (also called line-level) COVID-19 case and death data containing detailed information from individual case reports, CDC designed and implemented a new aggregate case surveillance system to inform emergency response decisions more efficiently, with timelier indicators of emerging areas of concern. We describe the processes implemented by CDC to operationalize this novel, multifaceted aggregate surveillance system for collecting COVID-19 case and death data to track the spread and impact of the SARS-CoV-2 virus at national, state, and county levels. We also review the processes established to acquire, process, and validate the aggregate number of cases and deaths due to COVID-19 in the United States at the county and jurisdiction levels during the pandemic. These processes include time-saving tools and strategies implemented to collect and validate authoritative COVID-19 case and death data from jurisdictions, such as web scraping to automate data collection and algorithms to identify and correct data anomalies. This topical review highlights the need to prepare for future emergencies, such as novel disease outbreaks, by having an event-agnostic aggregate surveillance system infrastructure in place to supplement line-level case reporting for near-real-time situational awareness and timely data.
Subject(s)
COVID-19 , Humans , United States/epidemiology , COVID-19/epidemiology , SARS-CoV-2 , Pandemics/prevention & control , Disease Outbreaks , Centers for Disease Control and Prevention, U.S.ABSTRACT
Machines powered by artificial intelligence increasingly mediate our social, cultural, economic and political interactions. Understanding the behaviour of artificial intelligence systems is essential to our ability to control their actions, reap their benefits and minimize their harms. Here we argue that this necessitates a broad scientific research agenda to study machine behaviour that incorporates and expands upon the discipline of computer science and includes insights from across the sciences. We first outline a set of questions that are fundamental to this emerging field and then explore the technical, legal and institutional constraints on the study of machine behaviour.
Subject(s)
Artificial Intelligence , Artificial Intelligence/legislation & jurisprudence , Artificial Intelligence/trends , Humans , Motivation , RoboticsABSTRACT
The field of artificial intelligence (AI) strives to build rational agents capable of perceiving the world around them and taking actions to advance specified goals. Put another way, AI researchers aim to construct a synthetic homo economicus, the mythical perfectly rational agent of neoclassical economics. We review progress toward creating this new species of machine, machina economicus, and discuss some challenges in designing AIs that can reason effectively in economic contexts. Supposing that AI succeeds in this quest, or at least comes close enough that it is useful to think about AIs in rationalistic terms, we ask how to design the rules of interaction in multi-agent systems that come to represent an economy of AIs. Theories of normative design from economics may prove more relevant for artificial agents than human agents, with AIs that better respect idealized assumptions of rationality than people, interacting through novel rules and incentive systems quite distinct from those tailored for people.
Subject(s)
Artificial Intelligence , Economics, Behavioral , Humans , ThinkingABSTRACT
To explain the spread of the 2014 Ebola epidemic in West Africa, and thus help with response planning, we analyzed publicly available data. We found that the risk for infection in an area can be predicted by case counts, population data, and distances between affected and nonaffected areas.
Subject(s)
Ebolavirus , Hemorrhagic Fever, Ebola/epidemiology , Hemorrhagic Fever, Ebola/transmission , Africa, Western/epidemiology , Geography, Medical , Humans , Models, Statistical , Population Surveillance , RiskABSTRACT
BACKGROUND: Schistosomiasis, a parasitic disease that affects over 200 million people, can lead to significant morbidity and mortality; distribution of single dose preventative chemotherapy significantly reduces disease burden. Implementation of control programs is dictated by disease prevalence rates, which are determined by costly and labor intensive screening of stool samples. Because ecological and human factors are known to contribute to the focal distribution of schistosomiasis, we sought to determine if specific environmental and geographic factors could be used to accurately predict Schistosoma mansoni prevalence in Nyanza Province, Kenya. METHODOLOGY/PRINCIPAL FINDINGS: A spatial mixed model was fit to assess associations with S. mansoni prevalence in schools. Data on S. mansoni prevalence and GPS location of the school were obtained from 457 primary schools. Environmental and geographic data layers were obtained from publicly available sources. Spatial models were constructed using ArcGIS 10 and R 2.13.0. Lower S.mansoni prevalence was associated with further distance (km) to Lake Victoria, higher day land surface temperature (LST), and higher monthly rainfall totals. Altitude, night LST, human influence index, normalized difference vegetation index, soil pH, soil texture, soil bulk density, soil water capacity, population, and land use variables were not significantly associated with S. mansoni prevalence. CONCLUSIONS: Our model suggests that there are specific environmental and geographic factors that influence S. mansoni prevalence rates in Nyanza Province, Kenya. Validation and use of schistosomiasis prevalence maps will allow control programs to plan and prioritize efficient control campaigns to decrease schistosomiasis burden.
Subject(s)
Geography , Models, Theoretical , Schistosoma mansoni/physiology , Schistosomiasis/epidemiology , Schistosomiasis/parasitology , Spatial Analysis , Animals , Child , Humans , Kenya/epidemiology , Multivariate Analysis , Prevalence , Regression AnalysisABSTRACT
Organisms, including Vibrio cholerae, can be transferred between harbors in the ballast water of ships. Zones in the Caribbean region where distance from shore and water depth meet International Maritime Organization guidelines for ballast water exchange are extremely limited. Use of ballast water treatment systems could mitigate the risk for organism transfer.
Subject(s)
Cholera Toxin/metabolism , Environmental Monitoring/methods , Seawater/microbiology , Ships , Vibrio cholerae/isolation & purification , Water Microbiology , Caribbean Region , Cholera/prevention & control , Cholera/transmission , DNA, Bacterial/genetics , Haiti , Vibrio cholerae/genetics , Vibrio cholerae/pathogenicity , Virulence , Waste Disposal, Fluid/methodsABSTRACT
During the 2010 cholera outbreak in Haiti, water and seafood samples were collected to detect Vibrio cholerae. The outbreak strain of toxigenic V. cholerae O1 serotype Ogawa was isolated from freshwater and seafood samples. The cholera toxin gene was detected in harbor water samples.