Hierarchical Object-oriented Spatio-Temporal Reasoning for Video Question Answering

Video Question Answering (Video QA) is a powerful testbed to develop new AI capabilities. This task necessitates learning to reason about objects, relations, and events across visual and linguistic domains in space-time. High-level reasoning demands lifting from associative visual pattern recognition to symbol like manipulation over objects, their behavior and interactions. Toward reaching this goal we propose an object-oriented reasoning approach in that video is abstracted as a dynamic stream of interacting objects. At each stage of the video event flow, these objects interact with each other, and their interactions are reasoned about with respect to the query and under the overall context of a video. This mechanism is materialized into a family of general-purpose neural units and their multi-level architecture called Hierarchical Object-oriented Spatio-Temporal Reasoning (HOSTR) networks. This neural model maintains the objects' consistent lifelines in the form of a hierarchically nested spatio-temporal graph. Within this graph, the dynamic interactive object-oriented representations are built up along the video sequence, hierarchically abstracted in a bottom-up manner, and converge toward the key information for the correct answer. The method is evaluated on multiple major Video QA datasets and establishes new state-of-the-arts in these tasks. Analysis into the model's behavior indicates that object-oriented reasoning is a reliable, interpretable and efficient approach to Video QA.

Few floods govern decades of suspended sediment flux in German upland rivers

<p>Sediment yield from lowland rivers around the globe is often dominated by suspended sediment that also acts as a carrier for pollutants and contaminants. Achieving a deeper understanding of the suspended sediment dynamics is important for river management, but often complicated by short or discontinuous time-series and scattered surveying locations. However, suspended sediment transport is highly variable in space and time, calling for decadal observations that reflect this variability. Here we make use of >130,000 measurements on water discharge (Q) and suspended sediment concentration (SSC) from twelve stations that drain large parts of the central German uplands, to investigate the spatiotemporal variability in suspended sediment flux. <br>The data has been collected during working days between 1965 and 2018 in context of the suspended sediment monitoring conducted by the Federal Waterways and Shipping Administration (WSV). The contributing catchments of the twelve monitoring stations range between 2500 and 22000 km<sup>2</sup> and cover observation periods between 27 and 53 years. <br>Despite roughly similar topographic and climatic conditions, average specific (suspended) sediment yield (SSY) varies between ~6 and ~29 t km<sup>2</sup> yr<sup>-1</sup>. Highest specific yields are observed for those catchments that drain the escarpment of the Swabian cuesta landscape. Even more pronounced than the spatial variability is the interannual variability in sediment yield, with SSY for very wet years exceeding SSY for dry years more than tenfold. Separating the hydrograph into base-flow and event-flow components, we find that sediment export during event-flows accounts for 60 to 85% of the long-term SSY, with individual floods accounting for more than 90% of the annual sediment export. We conclude that high specific (suspended) sediment yields in the central German uplands are conditioned by rapidly responding catchments (i.e. large fraction of event-flow contribution) with highly erodible lithologies of the Swabian cuesta landscapes.</p>

Simulation of non-stationary event flow with a nested stationary component

A method for constructing an ensemble of time series trajectories with a nonstationary flow of events and a non-stationary empirical distribution of the values of the observed random variable is described. We consider a special model that is similar in properties to some real processes, such as changes in the price of a financial instrument on the exchange. It is assumed that a random process is represented as an attachment of two processes - stationary and non-stationary. That is, the length of a series of elements in the sequence of the most likely event (the most likely price change in the sequence of transactions) forms a non-stationary time series, and the length of a series of other events is a stationary random process. It is considered that the flow of events is non-stationary Poisson process. A software package that solves the problem of modeling an ensemble of trajectories of an observed random variable is described. Both the values of a random variable and the time of occurrence of the event are modeled. An example of practical application of the model is given.

Simulation of non-stationary event flow with a nested stationary component

A method for constructing an ensemble of time series trajectories with a nonstationary flow of events and a non-stationary empirical distribution of the values of the observed random variable is described. We consider a special model that is similar in properties to some real processes, such as changes in the price of a financial instrument on the exchange. It is assumed that a random process is represented as an attachment of two processes - stationary and non-stationary. That is, the length of a series of elements in the sequence of the most likely event (the most likely price change in the sequence of transactions) forms a non-stationary time series, and the length of a series of other events is a stationary random process. It is considered that the flow of events is non-stationary Poisson process. A software package that solves the problem of modeling an ensemble of trajectories of an observed random variable is described. Both the values of a random variable and the time of occurrence of the event are modeled. An example of practical application of the model is given.

Simulation of non-stationary event flow with a nested stationary component

A method for constructing an ensemble of time series trajectories with a nonstationary flow of events and a non-stationary empirical distribution of the values of the observed random variable is described. We consider a special model that is similar in properties to some real processes, such as changes in the price of a financial instrument on the exchange. It is assumed that a random process is represented as an attachment of two processes - stationary and non-stationary. That is, the length of a series of elements in the sequence of the most likely event (the most likely price change in the sequence of transactions) forms a non-stationary time series, and the length of a series of other events is a stationary random process. It is considered that the flow of events is non-stationary Poisson process. A software package that solves the problem of modeling an ensemble of trajectories of an observed random variable is described. Both the values of a random variable and the time of occurrence of the event are modeled. An example of practical application of the model is given.

A multicentered academic medical center experience of a simulated root cause analysis (RCA) for hematology/oncology fellows.

188 Background: Quality improvement and patient safety education is an Accreditation Council for Graduate Medical Education (ACGME) common program requirement for hematology/oncology fellowships. Specifically, the ACGME requires trainee participation in interprofessional clinical patient safety activities, such as root cause analyses. These can be challenging to incorporate into busy schedules and are intimidating to some trainees, but simulated RCAs are a novel way to assure trainees gain important patient safety skills. We report on a multicentered experience utilizing a simulated RCA educational module in an attempt to provide fellows with the tools needed to participate in a live RCA and to increase awareness of the need to analyze patient safety events. Methods: The two-hour module included a didactic session explaining the basics of an RCA including common terminology, effective chart review, and personal interviews. The fellows assessed a patient safety event of a missed coagulopathy and created an event flow map and fishbone analysis. They then formed root cause/contributing factor statements and proposed a solution. Seventeen fellows from two institutions completed pre- and post-session surveys regarding the experience. Results: There was a 47% increase in both the percentage of fellows who felt comfortable participating in live RCAs in the future, and in the number of fellows who felt comfortable with using the tools typically utilized in an RCA. 70.59% of respondents felt that as a result of the mock RCA, they were more likely to report a near miss or adverse event. Conclusions: Mock RCAs are a feasible method of incorporating ACGME-required patient safety activities into hematology/oncology fellow education and are effective in increasing their comfort and understanding of important quality improvement skills