Search

In this paper we explore a time-depended SEIR model, in which the dynamics of the infection in four groups from a selected target group (population), divided according to the infection, are modeled by a system of nonlinear ordinary differential equations. Several basic parameters are involved in the model: coefficients of infection rate, incubation rate, recovery rate. The coefficients are adaptable to each specific infection, for each individual country, and depend on the measures to limit the spread of the infection and the effectiveness of the methods of treatment of the infected people in the respective country. If such coefficients are known, solving the nonlinear system is possible to be able to make some hypotheses for the development of the epidemic. This is the reason for using Bulgarian COVID-19 data to first of all, solve the so-called "inverse problem" and to find the parameters of the current situation. Reverse logic is initially used to determine the parameters of the model as a function of time, followed by computer solution of the problem. Namely, this means predicting the future behavior of these parameters, and finding (and as a consequence applying mass-scale measures, e.g., distancing, disinfection, limitation of public events), a suitable scenario for the change in the proportion of the numbers of the four studied groups in the future. In fact, based on these results we model the COVID-19 transmission dynamics in Bulgaria and make a two-week forecast for the numbers of new cases per day, active cases and recovered individuals. Such model, as we show, has been successful for prediction analysis in the Bulgarian situation. We also provide multiple examples of numerical experiments with visualization of the results., Comment: keywords: Nonlinear approximation, Mathematical modeling, data extrapolation

The broadening adoption of machine learning in the enterprise is increasing the pressure for strict governance and cost-effective performance, in particular for the common and consequential steps of model storage and inference. The RDBMS provides a natural starting point, given its mature infrastructure for fast data access and processing, along with support for enterprise features (e.g., encryption, auditing, high-availability). To take advantage of all of the above, we need to address a key concern: Can in-RDBMS scoring of ML models match (outperform?) the performance of dedicated frameworks? We answer the above positively by building Raven, a system that leverages native integration of ML runtimes (i.e., ONNX Runtime) deep within SQL Server, and a unified intermediate representation (IR) to enable advanced cross-optimizations between ML and DB operators. In this optimization space, we discover the most exciting research opportunities that combine DB/Compiler/ML thinking. Our initial evaluation on real data demonstrates performance gains of up to 5.5x from the native integration of ML in SQL Server, and up to 24x from cross-optimizations--we will demonstrate Raven live during the conference talk.