The last decade has seen a revolution in the theory and application of machine learning and pattern recognition. Through these advancements, variable ranking has emerged as an active and growing research area and it is now beginning to be applied to many new problems. The rationale behind this fact is that many pattern recognition problems are by nature ranking problems. The main objective of a ranking algorithm is to sort objects according to some criteria, so that, the most relevant items will appear early in the produced result list. Ranking methods can be analysed from two different methodological perspectives: ranking to learn and learning to rank. The former aims at studying methods and techniques to sort objects for improving the accuracy of a machine learning model. Enhancing a model performance can be challenging at times. For example, in pattern classification tasks, different data representations can complicate and hide the different explanatory factors of variation behind the data. In particular, hand-crafted features contain many cues that are either redundant or irrelevant, which turn out to reduce the overall accuracy of the classifier. In such a case feature selection is used, that, by producing ranked lists of features, helps to filter out the unwanted information. Moreover, in real-time systems (e.g., visual trackers) ranking approaches are used as optimization procedures which improve the robustness of the system that deals with the high variability of the image streams that change over time. The other way around, learning to rank is necessary in the construction of ranking models for information retrieval, biometric authentication, re-identification, and recommender systems. In this context, the ranking model's purpose is to sort objects according to their degrees of relevance, importance, or preference as defined in the specific application. This thesis addresses these issues and discusses different aspects of variable ranking in pattern recognition, biometrics, and computer vision. In particular, this work explores the merits of ranking to learn, by proposing novel solutions in feature selection that efficiently remove unwanted cues from the information stream. A novel graph-based ranking framework is proposed that exploits the convergence properties of power series of matrices thereby individuating candidate features, which turn out to be effective from a classification point of view. Moreover, it investigates the difficulties of ranking in real-time while presenting interesting solutions to better handle data variability in an important computer vision setting: Visual Object Tracking. The second part of this thesis focuses on the problem of learning to rank. Firstly, an interesting scenario of automatic user re-identification and verification in text chats is considered. Here, we start from the challenging problem of feature handcrafting to automatic feature learning solutions. We explore different techniques which turn out to produce effective ranks, contributing to push forward the state of the art. Moreover, we focus on advert recommendation, where deep convolutional neural networks with shallow architectures are used to rank ads according to users' preferences. We demonstrate the quality of our solutions in extensive experimental evaluations. Finally, this thesis introduces representative datasets and code libraries in different research areas that facilitate large-scale performance evaluation.
|Titolo:||Ranking to Learn and Learning to Rank: On the Role of Ranking in Pattern Recognition Applications|
|Data di pubblicazione:||2017|
|Appare nelle tipologie:||07.13 Doctoral Thesis|