Dipartimento di Ingegneria Informatica, Modellistica, Elettronica e Sistemistica - Tesi di Dottorato
Permanent URI for this collectionhttps://lisa.unical.it/handle/10955/31
Questa collezione raccoglie le Tesi di Dottorato afferenti al Dipartimento di Ingegneria Informatica, Modellistica, Elettronica e Sistemistica dell'Università della Calabria.
Browse
3 results
Search Results
Item An active learning Approach based on learning models' parameters exploitation(Università della Calabria, 2023-07-06) Scala, Francesco; Flesca, SergioArti cial Intelligence (AI) techniques and in particular Machine and Deep Learning (ML and DL), have been widely adopted to enhance various aspects of human life. ML algorithms can be categorized into four main types: Supervised Learning, Unsupervised Learning, Semi-supervised Learning, and Reinforcement Learning. A signi cant challenge in these techniques is the requirement for su cient labeled data for training. Active Learning (AL) is a machine learning framework that addresses this issue by selecting instances to be labeled in a smart way to optimize model training, i.e., AL reduces labeling time and leads to better-performing models by dynamically selecting the most representative samples to be labeled during the training phase. AL was proven to be e ective in di erent scenarios and its choice of querying a label depends on the cost and gain of obtaining the information. In this thesis, are presented two novel approaches for active learning in meta-learning models. The proposed methods, called LAL-IGradV and LAL-IGradV-VAE, select instances to be labeled using an estimate of their impact on the current classi er. This is achieved by evaluating the importance of previously labeled instances in training the classi cation model and training another model that estimates the importance of unlabeled instances. The approaches can be instantiated with any classi er that is trainable through gradient descent optimization, and in this study, is provided a formulation using a deep neural network. These approaches have not been thoroughly investigated in previous learning-to-active-learn methods and experimental results demonstrate its promising performance in scenarios where there are only a limited number of initially labeled instances. 2Item Methodologies and Applications for Big Data Analytics(Università della Calabria, 2020-05-02) Cassavia, Nunziato; Crupi, Felice; Flesca, Sergio; Masciari, ElioDue to the emerging Big Data paradigm, driven by the increase availability of users generated data, traditional data management techniques are inadequate in many real life scenarios. The availability of huge amounts of data pertaining to user social interactions calls for advanced analysis strategies in order to extract meaningful information. Furthermore, heterogeneity and high speed of user generated data require suitable data storage and management and a huge amount of computing power. This dissertation presents a Big Data framework able to enhances user quest for information by exploiting previous knowledge about their social environment. Moreover an introduction to Big Data and NoSQL systems is provided and two basic architecture for Big Data analysis are presented. The framework that enhances user quest, leverages the extent of influence that the users are potentially subject to and the influence they may exert on other users. User influence spread, across the network, is dynamically computed as well to improve user search strategy by providing specific suggestions, represented as tailored faceted features. The approach is tested in an important application scenario such as tourist recommendation where several experiment have been performed to assess system scalability and data read/write efficiency. The study of this system and of advanced analysis on Big Data has shown the need for a huge computing power. To this end an high performance computing system named CoremunitiTM is presented. This system represents a P2P solution for solving complex works by using the idling computational resources of users connected to this network. Users help each other by asking the network computational resources when they face high computing demanding tasks. Differently from many proposals available for volunteer computing, users providing their resources are rewarded with tangible credits. This approach is tested in an interesting scenario as 3D rendering where its efficiency has been compared with "traditional" commercial solutions like cloud platforms and render farms showing shorter task completion times at low cost.Item Methodologies and Applications for Big Data Analytics(Università della Calabria, 2020-05-02) Cassavia, Nunziato; Crupi, Felice; Flesca, Sergio; Masciari, Elio;Due to the emerging Big Data paradigm, driven by the increase availability of users generated data, traditional data management techniques are inadequate in many real life scenarios. The availability of huge amounts of data pertaining to user social interactions calls for advanced analysis strategies in order to extract meaningful information. Furthermore, heterogeneity and high speed of user generated data require suitable data storage and management and a huge amount of computing power. This dissertation presents a Big Data framework able to enhances user quest for information by exploiting previous knowledge about their social environment. Moreover an introduction to Big Data and NoSQL systems is provided and two basic architecture for Big Data analysis are presented. The framework that enhances user quest, leverages the extent of influence that the users are potentially subject to and the influence they may exert on other users. User influence spread, across the network, is dynamically computed as well to improve user search strategy by providing specific suggestions, represented as tailored faceted features. The approach is tested in an important application scenario such as tourist recommendation where several experiment have been performed to assess system scalability and data read/write efficiency. The study of this system and of advanced analysis on Big Data has shown the need for a huge computing power. To this end an high performance computing system named CoremunitiTM is presented. This system represents a P2P solution for solving complex works by using the idling computational resources of users connected to this network. Users help each other by asking the network computational resources when they face high computing demanding tasks. Differently from many proposals available for volunteer computing, users providing their resources are rewarded with tangible credits. This approach is tested in an interesting scenario as 3D rendering where its efficiency has been compared with "traditional" commercial solutions like cloud platforms and render farms showing shorter task completion times at low cost.