Machine Learning Algorithms for Detection of Shockable Rhythms in Automated Defibrillators

N. A. Alimov, V. A. Erofeeva, D. S. Shalymov

St. Petersburg State University

Institute of Problems of Mechanical Engineering RAS

Key words: pattern recognition, classification algorithms, machine learning, electrocardiograms, automated defibrillator.

Recently, the automation of a resuscitation process attracts more and more researchers due to the low level of survival rate in the cases of sudden cardiac arrest. Such condition mainly occurs when the hearts electrical activity becomes chaotic. In such cases it is required to deliver a shock by a defibrillator to restore the hearts normal activity as soon as possible. The defibrillators may be installed at public places, for example airports and supermarkets, to reduce the time before the resuscitation procedure begins. Nevertheless, the competence of a person delivering the shock may also be critical. Therefore, the automatic resuscitation procedure is more preferable. It is possible to apply machine learning methods to improve the survival rate. The goals of this paper are analyzing the existing classification methods and their validation on the real data, as well as the development of software for algorithm testing.

Bibliogr.: 23 refs.

Test Bench for Research of Adaptation Process of Planner with Feathers to the Changes in a Turbulent Airflow

Amelin K. S., Amelina N. O., Deryugin D. E., Ivanskiy Y. V.

St. Petersburg State University

Institute of Problems of Mechanical Engineering RAS {konstantinamelin, ngranichina, deryugin.denis, ivanskiy.yuriy}

Key words: fast gradient descent method, multiagent systems, transition process, airplane with feathers, turbulence.

The article considers the operation of the bench simulating the wing of an aircraft with movable elements (feathers), under conditions of turbulent air flow. By rotating the movable elements in two planes, depending on the characteristics of the air flow over the wing element, the effect of the chatter is reduced. By using the feedback mechanism between adjacent wing elements, a transient process is implemented. The transition happens from an unbalanced state in which the difference in pressure deviations from their values in the laminar flow differs greatly in different parts of the wing, to a finite number of clusters with the same pressure deviation, and then to synchronization of all wing elements and final equalization of the pressure difference. Synchronization is implemented by using a multi-agent algorithm based on the fast gradient method. In the article the mathematical model of a wing with moving elements is described, the problem formulation of synchronization of wing elements is presented, the synchronization algorithm is given, the element base of the used sensors is listed, the results of simulation experiments are analyzed.

Bibliogr.: 24 refs.

An Implementation of the Memory Manager in the Work-Stealing Load Balancer


Key words: optimal control, work-stealing deques, memory managers.

In this paper we describe the memory manager, implemented in the experimental work-stealing load balancer. The manager is based on the mathematical models of optimal control of work-stealing deques. The following methods of memory control are implemented in the manager and analyzed: optimal partition; partition in half. Also, the work of experimental and standard managers is compared.


Optimal Control of Two Queues, Working on the Principle of Custom Queues

A. M. Sazonov


Key words: data structures, custom queues, random walks, regular Markov chains.

In this paper we propose the mathematical model and a way of solving for the problem of optimal partitioning of shared memory for the custom queue. It is done in the form of two queues (the case of their sequential cyclic representation). It is also considered the imitation model for One after another representation.