We present a siamese adaptation of the Long Short-Term Memory (LSTM) network for labeled data comprised of pairs of variable-length sequences. Our model is applied to assess semantic similarity between sentences, where we exceed state of the art, outperforming carefully handcrafted features and recently proposed neural network systems of greater complexity. For these applications, we provide word-embedding vectors supplemented with synonymic information to the LSTMs, which use a fixed size vector to encode the underlying meaning expressed in a sentence (irrespective of the particular wording/syntax). By restricting subsequent operations to rely on a simple Manhattan metric, we compel the sentence representations learned by our model to form a highly structured space whose geometry reflects complex semantic relationships. Our results are the latest in a line of findings that showcase LSTMs as powerful language models capable of tasks requiring intricate understanding.

We address the problem of automatically acquiring knowledge of event sequences from text, with the aim of providing a predictive model for use in narrative generation systems. We present a neural network model that simultaneously learns embeddings for words describing events, a function to compose the embeddings into a representation of the event, and a coherence function to predict the strength of association between two events. We introduce a new development of the narrative cloze evaluation task, better suited to a setting where rich information about events is available. We compare models that learn vector-space representations of the events denoted by verbs in chains centering on a single protagonist. We find that recent work on learning vector-space embeddings to capture word meaning can be effectively applied to this task, including simple incorporation of a verb's arguments in the representation by vector addition. These representations provide a good initialization for learning the richer, compositional model of events with a neural network, vastly outperforming a number of baselines and competitive alternatives.

Evaluating the selection of content in a summary is important both for human-written summaries, which can be a useful pedagogical tool for reading and writing skills, and machine-generated summaries, which are increasingly being deployed in information management. The pyramid method assesses a summary by aggregating content units from the summaries of a wise crowd (a form of crowdsourcing). It has proven highly reliable but has largely depended on manual annotation. We propose PEAK, the first method to automatically assess summary content using the pyramid method that also generates the pyramid content models. PEAK relies on open information extraction and graph algorithms. The resulting scores correlate well with manually derived pyramid scores on both human and machine summaries, opening up the possibility of wide-spread use in numerous applications.

Semi-definite rank minimization problems model a wide range of applications in both signal processing and machine learning fields. This class of problem is NP-hard in general. In this paper, we propose a proximal Alternating Direction Method (ADM) for the well-known semi-definite rank regularized minimization problem. Specifically, we first reformulate this NP-hard problem as an equivalent biconvex MPEC (Mathematical Program with Equilibrium Constraints), and then solve it using proximal ADM, which involves solving a sequence of structured convex semi-definite subproblems to find a desirable solution to the original rank regularized optimization problem. Moreover, based on the Kurdyka-Lojasiewicz inequality, we prove that the proposed method always converges to a KKT stationary point under mild conditions. We apply the proposed method to the widely studied and popular sensor network localization problem. Our extensive experiments demonstrate that the proposed algorithm outperforms state-of-the-art low-rank semi-definite minimization algorithms in terms of solution quality.

In this paper, we focus on a novel knowledge reuse scenario where the knowledge in the source schema needs to be translated to a semantically heterogeneous target schema. We refer to this task as “knowledge translation” (KT). Unlike data translation and transfer learning, KT does not require any data from the source or target schema. We adopt a probabilistic approach to KT by representing the knowledge in the source schema, the mapping between the source and target schemas, and the resulting knowledge in the target schema all as probability distributions, specially using Markov random fields and Markov logic networks. Given the source knowledge and mappings, we use standard learning and inference algorithms for probabilistic graphical models to find an explicit probability distribution in the target schema that minimizes the Kullback-Leibler divergence from the implicit distribution. This gives us a compact probabilistic model that represents knowledge from the source schema as well as possible, respecting the uncertainty in both the source knowledge and the mapping. In experiments on both propositional and relational domains, we find that the knowledge obtained by KT is comparable to other approaches that require data, demonstrating that knowledge can be reused without data.

The Maximum Inner Product Search (MIPS) problem, prevalent in matrix factorization-based recommender systems, scales linearly with the number of objects to score. Recent work has shown that clever post-processing steps can turn the MIPS problem into a nearest neighbour one, allowing sublinear retrieval time either through Locality Sensitive Hashing or various tree structures that partition the Euclidian space. This work shows that instead of employing post-processing steps, substantially faster retrieval times can be achieved for the same accuracy when inference is not decoupled from the indexing process. By framing matrix factorization to be natively indexable, so that any solution is immediately sublinearly searchable, we use the machinery of Machine Learning to best learn such a solution. We introduce Indexable Probabilistic Matrix Factorization (IPMF) to shift the traditional post-processing complexity into the training phase of the model. Its inference procedure is based on Geodesic Monte Carlo, and adds minimal additional computational cost to standard Monte Carlo methods for matrix factorization. By coupling inference and indexing in this way, we achieve more than a 50% improvement in retrieval time against two state of the art methods, for a given level of accuracy in the recommendations of two large-scale recommender systems.

A construct that has been receiving attention recently in reinforcement learning is stochastic factorization (SF), a particular case of non-negative factorization (NMF) in which the matrices involved are stochastic. The idea is to use SF to approximate the transition matrices of a Markov decision process (MDP). This is useful for two reasons. First, learning the factors of the SF instead of the transition matrices can reduce significantly the number of parameters to be estimated. Second, it has been shown that SF can be used to reduce the number of operations needed to compute an MDP's value function. Recently, an algorithm called expectation-maximization SF (EMSF) has been proposed to compute a SF directly from transitions sampled from an MDP. In this paper we take a closer look at EMSF. First, by exploiting the assumptions underlying the algorithm, we show that it is possible to reduce it to simple multiplicative update rules similar to the ones that helped popularize NMF. Second, we analyze the optimization process underlying EMSF and find that it minimizes a modified version of the Kullback-Leibler divergence that is particularly well-suited for learning a SF from data sampled from an arbitrary distribution. Third, we build on this improved understanding of EMSF to draw an interesting connection with NMF and probabilistic latent semantic analysis. We also exploit the simplified update rules to introduce a new version of EMSF that generalizes and significantly improves its precursor. This new algorithm provides a practical mechanism to control the trade-off between memory usage and computing time, essentially freeing the space complexity of EMSF from its dependency on the number of sample transitions. The algorithm can also compute its approximation incrementally, which makes it possible to use it concomitantly with the collection of data. This feature makes the new version of EMSF particularly suitable for online reinforcement learning. Empirical results support the utility of the proposed algorithm.

When learning a hidden Markov model (HMM), sequential observations can often be complemented by real-valued summary response variables generated from the path of hidden states. Such settings arise in numerous domains, including many applications in biology, like motif discovery and genome annotation. In this paper, we present a flexible framework for jointly modeling both latent sequence features and the functional mapping that relates the summary response variables to the hidden state sequence. The algorithm is compatible with a rich set of mapping functions. Results show that the availability of additional continuous response variables can simultaneously improve the annotation of the sequential observations and yield good prediction performance in both synthetic data and real-world datasets.

Programming language processing (similar to natural language processing) is a hot research topic in the field of software engineering; it has also aroused growing interest in the artificial intelligence community. However, different from a natural language sentence, a program contains rich, explicit, and complicated structural information. Hence, traditional NLP models may be inappropriate for programs. In this paper, we propose a novel tree-based convolutional neural network (TBCNN) for programming language processing, in which a convolution kernel is designed over programs' abstract syntax trees to capture structural information. TBCNN is a generic architecture for programming language processing; our experiments show its effectiveness in two different program analysis tasks: classifying programs according to functionality, and detecting code snippets of certain patterns. TBCNN outperforms baseline methods, including several neural models for NLP.

Tagging has become increasingly important in many real-world applications noticeably including web applications, such as web blogs and resource sharing systems. Despite this importance, tagging methods often face difficult challenges such as limited training samples and incomplete labels, which usually lead to degenerated performance on tag prediction. To improve the generalization performance, in this paper, we propose Regularized Marginalized Cross-View learning (RMCV) by jointly modeling on attribute noise and label noise. In more details, the proposed model constructs infinite training examples with attribute noises from known exponential-family distributions and exploits label noise via marginalized denoising autoencoder. Therefore, the model benefits from its robustness and alleviates the problem of tag sparsity. While RMCV is a general method for learning tagging, in the evaluations we focus on the specific application of multi-label text tagging. Extensive evaluations on three benchmark data sets demonstrate that RMCV outstands with a superior performance in comparison with state-of-the-art methods.