Utjecaj entropije na točnost računa u impulsnom računalu

master's thesis

• Title (english): The influence of entropy on the accuracy of calculations in pulse computer
• Author: Mateja Batelić
• Mentor: Mario Stipčević (mentor)
• Mentor: Denis Sunko (komentor)
• Committee member: Denis Sunko (predsjednik povjerenstva)
• Committee member: Mario Stipčević (član povjerenstva)
• Committee member: Marinko Jablan (član povjerenstva)
• Committee member: Andrej Novak (član povjerenstva)
• Committee member: Mario Basletić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2022-09-28, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Impulsno neuronsko računanje predstavlja treću računalnu paradigmu zajedno sa digitalnim i kvantnim računalima, a zasniva se na biološki inspiriranom načinu procesuiranja informacija, odnosno na samim neuronima, živčanim stanicama živih bića. U ovom radu prezentirano je impulsno računanje sa svim dosad poznatim sklopovima za osnovne matematičke operacije – redom množenje, zbrajanje, dijeljenje i oduzimanje. Nadalje, objašnjene su različite formulacije entropije, pri čemu je stavljen naglasak na pojam entropijskog budžeta kao pouzdane motivacije za pronalaženje kvalitetnog determinističkog sklopa za zbrajanje i demonstraciju numeričke preciznosti u računima. U rezultatima su prezentirani poboljšani deterministički i nedeterministički sklopovi za zbrajanje, a pritom su provjeravani rezultati same operacije zbrajanja i vrijednosti relativne entropije. Povrh toga, proučavano je ponašanje najbolje ocjenjenih sklopova za zbrajanje u sklopu rješavanja kvadratne funkcije kombinirajući odabrane sklopove za zbrajanje sa sklopom za množenje, koji predstavlja najjednostavniji i najtočniji sklop u impulsnom računanju. Time je demonstrirana primjena sklopova u impulsnom računalu uz postizanje veće točnosti nakon provođenja matematičkih operacija kroz kombinaciju naizgled konceptualno jednostavnih sklopova.

Abstract (english)

Neuronal pulse computing represents third paradigm along with digital and quantum computing, being based on biologically-inspired way of processing information, i.e., on neurons, which are neural cells in the body of living beings. In this work, pulse computing is presented together with all known circuits so far, for basic mathematical operations – respectively multiplication, addition, division, and subtraction. Furthermore, different formulations of entropy are explained, whereby emphasis is being placed on the concept of entropy budget as an underlying motivation for finding a high-quality deterministic circuit for addition and for demonstration of numerical precisions in calculations. In the results section, we present improved deterministic and nondeterministic addition circuits. In particular, the results of the addition operation itself and relative entropy values were checked. On top of that, the behaviour of the best rated addition circuits was studied to solve quadratic function by combining the selected circuits with the multiplication circuit, which represents the simplest and the most accurate circuit in pulse computing. This demonstrated the application of circuits in pulse computing, together with achieving higher precision, after conducting mathematical operations through seemingly conceptually simple circuits.

• Keywords: impulsno računanje; entropijski budžet; relativna entropija
• Keywords (english): pulse computing; entropy budget; relative entropy
• Language: croatian
• URN:NBN: urn:nbn:hr:217:167082
• Study programme: Title: Physics; specializations in: Research; Course: Research; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra fizike (magistar/magistra fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-10-18 08:27:54