The rapid construction of supervised text classification models is becoming a pervasive need across many modern applications. To reduce human-labeling bottlenecks, many new statistical paradigms (e.g., active, semi-supervised, transfer and multi-task learning) have been vigorously pursued in recent literature with varying degrees of empirical success. Concurrently, the emergence of Web 2.0 platforms in the last decade has enabled a world-wide, collaborative human effort to construct a massive ontology of concepts with very rich, detailed and accurate descriptions. In this paper we propose a new framework to extract supervisory information from such ontologies and complement it with a shift in human effort from direct labeling of examples in the domain of interest to the much more efficient identification of concept-class associations. Through empirical studies on text categorization problems using the Wikipedia ontology, we show that this shift allows very high-quality models to be immediately induced at virtually no cost.

In recent years, the summarisation and decomposition of social networks has become increasingly popular, from community finding to role equivalence. However, these approaches concentrate on one type of model only. Generalised block modelling decomposes a network into independent, interpretable, labeled blocks, where the block labels summarise the relationship between two sets of users. Existing algorithms for fitting generalised block models do not scale beyond networks of 100 vertices. In this paper, we introduce two new algorithms, one based on genetic algorithms and the other on simulated annealing, that is at least two orders of magnitude faster than existing algorithms and obtaining similar accuracy. Using synthetic and real datasets, we demonstrate their efficiency and accuracy and show how generalised block modelling and our new approaches enable tractable network summarisation and modelling of medium sized networks.

Another way to speed up a RL algorithm is by using Transfer Learning, a paradigm of machine learning that In this paper we propose to combine three AI techniques reuses knowledge accumulated in a previous task to speed up to speed up a Reinforcement Learning algorithm the learning of a novel, but related, target task [Taylor and in a Transfer Learning problem: Casebased Stone, 2009]. Reasoning, Heuristically Accelerated Reinforcement This paper investigates the use of the Case-Based Heuristically Learning and Neural Networks. To do Accelerated Reinforcement Learning (CB-HARL) algorithm so, we propose a new algorithm, called L3, which [Bianchi et al., 2009] as a means to transfer learning works in 3 stages: in the first stage, it uses Reinforcement acquired by one agent during its training in one problem to Learning to learn how to perform one another agent that has to learn how to solve a similar, but task, and stores the optimal policy for this problem more complex, problem. To do so, we propose a new algorithm, as a case-base; in the second stage, it uses a Neural called L3, which works in 3 stages: in the first stage, Network to map actions from one domain to actions it uses the Q-learning algorithm [Watkins, 1989] to learn how in the other domain and; in the third stage, it uses to perform one task, and stores the optimal policy for this the case-base learned in the first stage as heuristics problem as a case-base; in the second stage, it uses a Neural to speed up the learning performance in a related, Network to map actions from one domain to actions in but different, task. The RL algorithm used the other domain and; in the third stage, it uses the case-base in the first phase is the Q-learning and in the third learned in the first stage as heuristics in the CB-HARL algorithm, phase is the recently proposed Case-based Heuristically speeding up the learning process.

Topic models have a wide range of applications, including modeling of text documents, images, user preferences, product rankings, and many others. However, learning optimal models may be difficult, especially for large problems. The reason is that inference techniques such as Gibbs sampling often converge to suboptimal models due to the abundance of local minima in large datasets. In this paper, we propose a general method of improving the performance of topic models. The method, called 'grouping transform', works by introducing auxiliary variables which represent assignments of the original model tokens to groups. Using these auxiliary variables, it becomes possible to resample an entire group of tokens at a time. This allows the sampler to make larger state space moves. As a result, better models are learned and performance is improved. The proposed ideas are illustrated on several topic models and several text and image datasets. We show that the grouping transform significantly improves performance over standard models.

In this paper, we introduce a probabilistic classification model to address the task of semi-supervised learning. The major novelty of our proposal stems from measuring distributional relationships between the labeled and unlabeled data. This is achieved from a stochastic translation model between data distributions that is estimated from a mixture model. The proposed classifier is defined from the combination of both the translation model and a kernel logistic regression on labeled data. Experimental results obtained over synthetic and real-world data sets validate the usefulness of our proposal.

In this paper, we propose a new framework for constructing text metrics which can be used to compare and support inferences among terms and sets of terms. Our metric is derived from data-driven kernels on graphs that let us capture global relations among terms and sets of terms, regardless of their complexity and size. To compute the metric efficiently for any two subsets of terms, we develop an approximation technique that relies on the precompiled term-term similarities. To scale-up the approach to problems with huge number of terms, we develop and experiment with a solution that subsamples the term space. We demonstrate the benefits of the whole framework on two text inference tasks: prediction of terms in the article from its abstract and query expansion in information retrieval.

In this work we propose an approach for generalization in continuous domain Reinforcement Learning that, instead of using a single function approximator, tries many different function approximators in parallel, each one defined in a different region of the domain. Associated with each approximator is a relevance function that locally quantifies the quality of its approximation, so that, at each input point, the approximator with highest relevance can be selected. The relevance function is defined using parametric estimations of the variance of the q-values and the density of samples in the input space, which are used to quantify the accuracy and the confidence in the approximation, respectively. These parametric estimations are obtained from a probability density distribution represented as a Gaussian Mixture Model embedded in the input-output space of each approximator. In our experiments, the proposed approach required a lesser number of experiences for learning and produced more stable convergence profiles than when using a single function approximator.

Uncertainty and vagueness are pervasive phenomena in real-life knowledge. They are supported in extended description logics that adapt classical description logics to deal with numerical probabilities or fuzzy truth degrees. While the two concepts are distinguished for good reasons, they combine in the notion of probably, which is ultimately a fuzzy qualification of probabilities. Here, we develop existing propositional logics of fuzzy probability into a full-blown description logic, and we show decidability of several variants of this logic under Lukasiewicz semantics. We obtain these results in a novel generic framework of fuzzy coalgebraic logic; this enables us to extend our results to logics that combine crisp ingredients including standard crisp roles and crisp numerical probabilities with fuzzy roles and fuzzy probabilities.

This paper studies a computational logic for dishonest reasoning. We introduce logic programs with disinformation to represent and reason with dishonesty. We then consider two different cases of dishonesty: deductive dishonesty and abductive dishonesty. The former misleads another agent to deduce wrong conclusions, while the latter interrupts another agent to abduce correct explanations. In deductive or abductive dishonesty, an agent can perform different types of dishonest reasoning such as lying, bullshitting, and withholding information. We show that these different types of dishonest reasoning are characterized by extended abduction, and address their computational methods using abductive logic programming.

We present a new and compelling approach to the efficient solution of important computational problems that arise in the context of abstract argumentation. Our approach makes known algorithms defined for restricted fragments generally applicable, at a computational cost that scales with the distance from the fragment. Thus, in a certain sense, we gradually augment tractable fragments. Surprisingly, it turns out that some tractable fragments admit such an augmentation and that others do not. More specifically, we show that the problems of credulous and skeptical acceptance are fixed-parameter tractable when parameterized by the distance from the fragment of acyclic argumentation frameworks. Other tractable fragments such as the fragments of symmetrical and bipartite frameworks seem to prohibit an augmentation: the acceptance problems are already intractable for frameworks at distance 1 from the fragments. For our study we use a broad setting and consider several different semantics. For the algorithmic results we utilize recent advances in fixed-parameter tractability.