Anton Gunzinger

All-night sleep EEG and artificial stochastic control signals have similar correlation dimensions ☆

EEG signals have been considered to be generated either by stochastic processes or by non-linear deterministic systems exhibiting chaotic behavior. To address this problem, the correlation dimension of the EEG was computed and compared to the correlation dimension of an artificial signal with identical power spectrum. By using a new type of personal super computer we were able for the first time to calculate the correlation dimension for the sleep episode of an entire night as well as for the corresponding artificial signal. The correlation dimension was high in episodes of rapid eye movement (REM) sleep, declined progressively within each non-REM sleep episode, and reached a low level at times when EEG slow waves (0.75-4.5 Hz) were dominant. The correlation dimension of the artificial signal and the EEG changed largely in parallel, although on average the values of the artificial signal were 7.3% higher. These results do not support the hypothesis that the sleep EEG is generated by a chaotic attractor.

Correlation Dimension of the Human Sleep Electroencephalogram: Cyclic Changes in the Course of the Night

The complexity of the electroencephalogram (EEG) during human sleep can be estimated by calculating the correlation dimension. Due to the large number of calculations required by this approach, only selected short (4–164 s) segments of the sleep EEG have been analysed previously. By using a new type of personal supercomputer, we were able to calculate the correlation dimension of overlapping 1 min EEG segments for the entire sleep episode (480 min) of 11 subjects and thereby delineate the time course of the changes. The correlation dimension was high in episodes of rapid eye movement (REM) sleep, declined progressively within each non‐REM sleep episode, and reached a low level at times when EEG slow waves (0.75–4.5 Hz) were dominant. However, whereas slow‐wave activity showed its typical progressive decline from non‐REMIREM sleep cycle 1 to 4, no such trend was present for the correlation dimension. By providing an estimate of the complexity of a signal and being independent of amplitude and frequency measures, the correlation dimension represents a novel approach to exploring the dynamics of sleep and the processes underlying its regulation.

Architecture and implementation of a single-board desktop supercomputer

A distributed-memory parallel computer (the “α7”) with a peak-performance of more than 1 GPlop was implemented on an area of 300 square inch. The seven processing elements are interconnected using Intelligent Communication — a communication scheme already approved in the MUSIC, a DSP based parallel computer of the Electronics Lab. The combination of processing elements built up using standard components with a fast, low-latency communication scheme implemented in hardware leads to a powerful parallel computer for the scientists or engineers workplace. Several of these computers can be joined via an optical network to form a distributed supercomputer.

Parallel molecular dynamics on a multi signalprocessor system

Abstract This paper gives an overview of a parallel computer architecture called MUSIC (Multi Signalprocessor System with Intelligent Communication), which has been developed at the Swiss Federal Institute of Technology. The current version achieves a peak performance of 3.8 GFlops. We discuss the system software and tools used to program the system and then present our implementation of a molecular dynamics simulation program which uses the architecture of MUSIC in an efficient way. We demonstrate the correctness of our implementation and give measurements of the performance of the system. To the best of our knowledge, MUSIC outperforms the most powerful present-day vector supercomputers.

Estimation of the Correlation Dimension of All-Night Sleep EEG Data with a Personal Super Computer

The correlation dimension (CD) is a gauge of the complexity of the recorded signal. An algorithm proposed by (1985) was implemented on the personal super computer MUSIC (multi processor system with intelligent communication) which allowed to analyze for the first time the data of entire sleep episodes. The all-night sleep EEG of a subject recorded under three different conditions (baseline (BL), sleep in a sitting condition after intake of placebo (PL) or triazolam (TR)) served as the data base. The CD of the EEG was modulated by the sleep cycle. Median values decreased from nonREM (non rapid eye movement) sleep stage 1 to 4. For REM (rapid eye movement) sleep the values were between those of stage 1 and 2. For sleep in a sitting position, the CD in REM sleep (PL and TR) and stage 2 (PL) was increased compared with BL. In stage 4 the CD was lower for TR than for BL and PL, and higher for PL than BL.

The SYnchronous DAtaflow MAchine: Architecture and Performance

In this paper a parallel computer architecture for real time image processing is described. The architecture centers on the direct mapping of a static dataflow graph into hardware: each node (or group of nodes) is replaced by a processing element.

Datenflußrechner zur Echtzeitbildverarbeitung: Softwareentwicklungsumgebung

Viele leistungsfahige Hardwarearchitekturen zur Bildverarbeitung sind bekannt /1,2,3/, doch die Implementierung neuer, vom Entwickler nicht vorbereiteter Algorithmen verursacht oft unuberwindbare Schwierigkeiten; sei es, dass die Anzahl der gleichzeitig arbeitenden Hardwaremodule durch das Systemkonzept bedingt relativ niedrig bleibt, sei es, dass die Implementierung neuer Algorithmen auf “tiefstem Hardwareniveau” erfolgen muss.

New scanning approach for motion picture digitizing

For digital restoration of motion picture films the correct scanning (actually measuring physical properties i.e. optical densities) is vital. Therefore thus, the delicate handling of aged films and the prohibitive high prices of commercially available scanners forced the development of a simple yet good quality film scanning device. The scanning principle and the increasing aspects specific to digital restoration will be discussed.

Programming Environment for a High-Performance Parallel Supercomputer with Intelligent Communication

At the Electronics Laboratory of the Swiss Federal Institute of Technology (ETH) in Zurich, the high-performance parallel supercomputer MUSIC (MUlti processor System with Intelligent Communication) has been developed. As applications like neural network simulation and molecular dynamics show, the Electronics Laboratory supercomputer is absolutely on par with those of conventional supercomputers, but electric power requirements are reduced by a factor of 1,000, weight is reduced by a factor of 400, and price is reduced by a factor of 100. Software development is a key issue of such parallel systems. This article focuses on the programming environment of the MUSIC system and on its applications.

Dataparallel Programming with Intelligent Communication

A key feature in data-parallel programming is the potential to distribute and redistribute data among the processing elements (PEs) efficiently. Dataparallel programming is very well suited for parallel systems with distributed memory. The execution of a single program (SPMD) on different data partitions is easy to implement and often results in a performance close to optimal. A problem arises only when data has to be exchanged among the PEs. Normally distribution is done by sending messages (between PEs) or through virtual shared-memory. The cost for this communication limits the performance of parallel systems. The intelligent communication (IC) scheme delivers a powerful feature to the programmer, which makes distribution and redistribution of data simple and with hardware support extremely efficient. This paper describes the IC scheme and shows the advantages over traditional MP systems. Also two hardware implementations and several possible mappings on MP systems are shown. Performance results give a closer view of the superiority of intelligent communication.