Visual object recognition is one of the key human capabilities that we would like machines to have. The problem is the following: given an image of an object (e.g. someone's face), predict its label (e.g. that person's name) from a set of possible object labels. The predominant approach to solving the recognition problem has been to learn a discriminative model, i.e. a model of the conditional probability $P(l|v)$ over possible object labels $l$ given an image $v$. Here we consider an alternative class of models, broadly referred to as \emph{generative models}, that learns the latent structure of the image so as to explain how it was generated. This is in contrast to discriminative models, which dedicate their parameters exclusively to representing the conditional distribution $P(l|v)$. Making finer distinctions among generative models, we consider a supervised generative model of the joint distribution $P(v,l)$ over image-label pairs, an unsupervised generative model of the distribution $P(v)$ over images alone, and an unsupervised \emph{reconstructive} model, which includes models such as autoencoders that can reconstruct a given image, but do not define a proper distribution over images. The goal of this thesis is to empirically demonstrate various ways of using these models for object recognition. Its main conclusion is that such models are not only useful for recognition, but can even outperform purely discriminative models on difficult recognition tasks. We explore four types of applications of generative/reconstructive models for recognition: 1) incorporating complex domain knowledge into the learning by inverting a synthesis model, 2) using the latent image representations of generative/reconstructive models for recognition, 3) optimizing a hybrid generative-discriminative loss function, and 4) creating additional synthetic data for training more accurate discriminative models. Taken together, the results for these applications support the idea that generative/reconstructive models and unsupervised learning have a key role to play in building object recognition systems.

Online Social Networks (OSNs) have seen an enormous boost in popularity in recent years. Along with this popularity has come tribulations such as privacy concerns, spam, phishing and malware. Many recent works have focused on automatically detecting these unwanted behaviors in OSNs so that they may be removed. These works have developed state-of-the-art detection schemes that use machine learning techniques to automatically classify OSN accounts as spam or non-spam. In this work, these detection schemes are recreated and tested on new data. Through this analysis, it is clear that spammers are beginning to evade even these detectors. The evasion tactics used by spammers are identified and analyzed. Then a new detection scheme is built upon the previous ones that is robust against these evasion tactics. Next, the difficulty of evasion of the existing detectors and the new detector are formalized and compared. This work builds a foundation for future researchers to build on so that those who would like to protect innocent internet users from spam and malicious content can overcome the advances of those that would prey on these users for a meager dollar.

Thesis (M.S.)--University of Toledo, 2009. / Typescript. "Submitted as partial fulfillment of the requirements for The Master of Science in Engineering." "A thesis entitled"--at head of title. Bibliography: leaves 158-164.

Thesis (Ph. D.)--Oregon State University, 2010. / Printout. Includes bibliographical references (leaves 122-126). Also available on the World Wide Web.

Many important real-world robotic tasks have high diameter, that is, their solution requires a large number of primitive actions by the robot. For example, they may require navigating to distant locations using primitive motor control commands. In addition, modern robots are endowed with rich, high-dimensional sensory systems, providing measurements of a continuous environment. Reinforcement learning (RL) has shown promise as a method for automatic learning of robot behavior, but current methods work best on lowdiameter, low-dimensional tasks. Because of this problem, the success of RL on real-world tasks still depends on human analysis of the robot, environment, and task to provide a useful set of perceptual features and an appropriate decomposition of the task into subtasks. This thesis presents Self-Organizing Distinctive-state Abstraction (SODA) as a solution to this problem. Using SODA a robot with little prior knowledge of its sensorimotor system, environment, and task can automatically reduce the effective diameter of its tasks. First it uses a self-organizing feature map to learn higher level perceptual features while exploring using primitive, local actions. Then, using the learned features as input, it learns a set of high-level actions that carry the robot between perceptually distinctive states in the environment. Experiments in two robot navigation environments demonstrate that SODA learns useful features and high-level actions, that using these new actions dramatically speeds up learning for high-diameter navigation tasks, and that the method scales to large (buildingsized) robot environments. These experiments demonstrate SODAs effectiveness as a generic learning agent for mobile robot navigation, pointing the way toward developmental robots that learn to understand themselves and their environments through experience in the world, reducing the need for human engineering for each new robotic application. / text

Not available / text

Not available

Email has become a part of most people's lives, and the ever increasing amount of messages people receive can lead to email overload. We attempt to mitigate this problem using email thread summarization. Summaries can be used for things other than just replacing an incoming email message. They can be used in the business world as a form of corporate memory, or to allow a new team member an easy way to catch up on an ongoing conversation. Email threads are of particular interest to summarization because they contain much structural redundancy due to their conversational nature. Our email thread summarization approach uses machine learning to pick which sentences from the email thread to use in the summary. A machine learning summarizer must be trained using previously labeled data, i.e. manually created summaries. After being trained our summarization algorithm can generate summaries that on average contain over 70% of the same sentences as human annotators. We show that labeling some key features such as speech acts, meta sentences, and subjectivity can improve performance to over 80% weighted recall. To create such email summarization software, an email dataset is needed for training and evaluation. Since email communication is a private matter, it is hard to get access to real emails for research. Furthermore these emails must be annotated with human generated summaries as well. As these annotated datasets are rare, we have created one and made it publicly available. The BC3 corpus contains annotations for 40 email threads which include extractive summaries, abstractive summaries with links, and labeled speech acts, meta sentences, and subjective sentences. While previous research has shown that machine learning algorithms are a promising approach to email summarization, there has not been a study on the impact of the choice of algorithm. We explore new techniques in email thread summarization using several different kinds of regression, and the results show that the choice of classifier is very critical. We also present a novel feature set for email summarization and do analysis on two email corpora: the BC3 corpus and the Enron corpus.

No description available.

We explore the training of deep neural networks to produce vector representations using weakly labelled information in the form of binary similarity labels for pairs of training images. Previous methods such as siamese networks, IMAX and others, have used fixed cost functions such as $L_1$, $L_2$-norms and mutual information to drive the representations of similar images together and different images apart. In this work, we formulate learning as maximizing the likelihood of binary similarity labels for pairs of input images, under a parameterized probabilistic similarity model. We describe and evaluate several forms of the similarity model that account for false positives and false negatives differently. We extract representations of MNIST, AT\&T ORL and COIL-100 images and use them to obtain classification results. We compare these results with state-of-the-art techniques such as deep neural networks and convolutional neural networks. We also study our method from a dimensionality reduction prospective.