In a specular way of what happens in our life experiences, also the world of engineering and computer science is subject to numerous causes of uncertainty. They may alter the desired behaviour of algorithms, systems (more or less complex) and practical applications. In this work, after a first presentation and investigation phase concerning the phenomenon of uncertainty, with suitable definitions and taxonomies, we will discuss about uncertainty and related problems. We will deal in particular with uncertainty and its management in the context of self-adaptive systems. Knowing the runtime behaviour of a system during its execution can help to understand if is the case to change running configuration (in terms of self-adaptiveness) in order to achieve better results. A lot of malfunctioning or bad performance causes can be related to various sources of uncertainty (and their variables), it should not be a surprise for us. This is the starting point of this work: knowing that uncertainty affects our system, try to face off the linked problems with appropriate paradigms imported from the field of AI (Artificial Intelligence). In particular, we will adopt methodologies and approaches regarding Fuzzy Logic and ANN (Artificial Neural Networks). Before starting the experimentation phase with these methodologies, we will analyse some alternatives. Some of these alternatives are imported from other fields and adapted as possible. The real world application under study will be presented in [Chapter 3]; all data adopted in this work are public domain data, properly extracted from real world observations in (servers’) logs. We will start from this data to extract knowledge concerning the application under analysis, in order to possibly improve the self- adaptive module. In this way we explore both the theoretical and the practical aspects regarding the themes of this work.
In maniera speculare a quanto succede per le nostre esperienze di vita, anche il mondo dell’ingegneria e dell’informatica è soggetto a numerose cause di incertezza che possono alterare il normale corso di algoritmi, sistemi (più o meno complessi) ed applicazioni pratiche. In questo elaborato, dopo aver presentato ed investigato in merito al fenomeno dell’incertezza, con opportune definizioni e tassonomie, ci occuperemo della gestione dell’incertezza stessa, con un occhio di riguardo nel contesto dei sistemi auto-adattivi. Conoscere il comportamento di un dato sistema durante la sua esecuzione può farci comprendere che bisogna optare per una diversa configurazione (in termini auto-adattivi) per ottenere migliori risultati. Non c’è da stupirsi se tra le possibili cause relative a malfunzionamenti o performance non in linea con gli standard ci sono proprio variabile legate alle varie forme di incertezza. Da qui sorge l’esigenza di questo elaborato: ovvero far fronte alle cause di incertezza adoperando paradigmi importati dal campo dell’Intelligenza Artificiale. Nello specifico adopereremo metodologie ed approcci inerenti la Logica Fuzzy e le Reti Neurali. Prima di avventurarci nel campo sperimentale con i metodi sopra citati, analizzeremo alcune alternative. Tra queste figurano interessanti alternative importate da altri campi ed adattate ad hoc. L’applicazione del mondo reale che sarà oggetto di studio verrà descritta ed analizzata nell’apposito [Capitolo 3]; tutti i dati che verranno adoperati sono di dominio pubblico e riguardano osservazioni reali estratte opportunamente da log (quindi questi ultimi non sono frutto di simulazioni o generazioni casuali). Questo modus operandi è stato stabilito per l’abbondante disponibilità dei dati e per avere un maggiore contatto con il mondo reale, in modo tale da toccare con mano l’influenza dell’incertezza non solo nella teoria, ma anche nella pratica.
Dealing with uncertainties in availability models using fuzzy logic and neural networks
MARCHESANI, FRANCESCO
2016/2017
Abstract
In a specular way of what happens in our life experiences, also the world of engineering and computer science is subject to numerous causes of uncertainty. They may alter the desired behaviour of algorithms, systems (more or less complex) and practical applications. In this work, after a first presentation and investigation phase concerning the phenomenon of uncertainty, with suitable definitions and taxonomies, we will discuss about uncertainty and related problems. We will deal in particular with uncertainty and its management in the context of self-adaptive systems. Knowing the runtime behaviour of a system during its execution can help to understand if is the case to change running configuration (in terms of self-adaptiveness) in order to achieve better results. A lot of malfunctioning or bad performance causes can be related to various sources of uncertainty (and their variables), it should not be a surprise for us. This is the starting point of this work: knowing that uncertainty affects our system, try to face off the linked problems with appropriate paradigms imported from the field of AI (Artificial Intelligence). In particular, we will adopt methodologies and approaches regarding Fuzzy Logic and ANN (Artificial Neural Networks). Before starting the experimentation phase with these methodologies, we will analyse some alternatives. Some of these alternatives are imported from other fields and adapted as possible. The real world application under study will be presented in [Chapter 3]; all data adopted in this work are public domain data, properly extracted from real world observations in (servers’) logs. We will start from this data to extract knowledge concerning the application under analysis, in order to possibly improve the self- adaptive module. In this way we explore both the theoretical and the practical aspects regarding the themes of this work.File | Dimensione | Formato | |
---|---|---|---|
2017_12_Marchesani.pdf
accessibile in internet per tutti
Descrizione: Testo della tesi di laurea magistrale (aggiornato)
Dimensione
16.24 MB
Formato
Adobe PDF
|
16.24 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/137692