It is crucial yet challenging for deep learning models to properly characterize uncertainty that is pervasive in real-world environments. Heteroscedastic neural networks (HNNs) are promising methods that capture data uncertainty for forecasting problems while existing HNNs have difficulties in conjoining calibrated uncertainty estimation and satisfactory predictive performance due to the failure to construct an explicit interaction between the prediction and its associated uncertainty. This article develops SDE-HNN, an improved HNN equipped with stochastic differential equations (SDE), to characterize the interaction between the predictive mean and variance inside HNNs for accurate and reliable forecasting. The existence and uniqueness of the solution to the devised neural SDE are guaranteed. Moreover, based on the bias-variance tradeoff for the optimization in SDE-HNN, we design an enhanced numerical SDE solver to improve learning stability. Finally, we present two new diagnostic uncertainty metrics to systematically evaluate the predictive uncertainty. Experiments on various challenging datasets show that our method significantly outperforms state-of-the-art baselines on both predictive performance and uncertainty quantification, delivering well-calibrated and sharp prediction intervals in time-series forecasting.

Partial differential equations (PDEs) that fit scientific data can represent physical laws with explainable mechanisms for various mathematically-oriented subjects, such as physics and finance. The data-driven discovery of PDEs from scientific data thrives as a new attempt to model complex phenomena in nature, but the effectiveness of current practice is typically limited by the scarcity of data and the complexity of phenomena. Especially, the discovery of PDEs with highly nonlinear coefficients from low-quality data remains largely under-addressed. To deal with this challenge, we propose a novel physics-guided learning method, which can not only encode observation knowledge such as initial and boundary conditions but also incorporate the basic physical principles and laws to guide the model optimization. We theoretically show that our proposed method strictly reduces the coefficient estimation error of existing baselines, and is also robust against noise. Extensive experiments show that the proposed method is more robust against data noise, and can reduce the estimation error by a large margin. Moreover, all the PDEs in the experiments are correctly discovered, and for the first time we are able to discover three-dimensional PDEs with highly nonlinear coefficients.

For reliability, machine learning models in some areas, e.g., finance and healthcare, require to be both accurate and globally interpretable. Among them, credit risk assessment is a major application of machine learning for financial institutions to evaluate credit of users and detect default or fraud. Simple white-box models, such as Logistic Regression (LR), are usually used for credit risk assessment, but not powerful enough to model complex nonlinear interactions among features. In contrast, complex black-box models are powerful at modeling, but lack of interpretability, especially global interpretability. Fortunately, automatic feature crossing is a promising way to find cross features to make simple classifiers to be more accurate without heavy handcrafted feature engineering. However, existing automatic feature crossing methods have problems in efficiency on credit risk assessment, for corresponding data usually contains hundreds of feature fields.

Large-scale topic models serve as basic tools for feature extraction and dimensionality reduction in many practical applications. As a natural extension of flat topic models, hierarchical topic models (HTMs) are able to learn topics of different levels of abstraction, which lead to deeper understanding and better generalization than their flat counterparts. However, existing scalable systems for flat topic models cannot handle HTMs, due to their complicated data structures such as trees and concurrent dynamically growing matrices, as well as their susceptibility to local optima. In this paper, we study the hierarchical latent Dirichlet allocation (hLDA) model which is a powerful nonparametric Bayesian HTM. We propose an efficient partially collapsed Gibbs sampling algorithm for hLDA, as well as an initialization strategy to deal with local optima introduced by tree-structured models. We also identify new system challenges in building scalable systems for HTMs, and propose efficient data layout for vectorizing HTM as well as distributed data structures including dynamic matrices and trees. Empirical studies show that our system is 87 times more efficient than the previous open-source implementation for hLDA, and can scale to thousands of CPU cores. We demonstrate our scalability on a 131-million-document corpus with 28 billion tokens, which is 4--5 orders of magnitude larger than previously used corpus. Our distributed implementation can extract 1,722 topics from the corpus with 50 machines in just 7 hours.

Learning large-scale Latent Dirichlet Allocation (LDA) models is beneficial for many applications that involve large collections of documents.Recent work has been focusing on developing distributed algorithms in the batch setting, while leaving stochastic methods behind, which can effectively explore statistical redundancy in big data and thereby are complementary to distributed computing.The distributed stochastic gradient Langevin dynamics (DSGLD) represents one attempt to combine stochastic sampling and distributed computing, but it suffers from drawbacks such as excessive communications and sensitivity to partitioning of datasets across nodes. DSGLD is typically limited to learn small models that have about 103 topics and $10^3$ vocabulary size.

Dynamic topic models (DTMs) are very effective in discovering topics and capturing their evolution trends in time series data. To do posterior inference of DTMs, existing methods are all batch algorithms that scan the full dataset before each update of the model and make inexact variational approximations with mean-field assumptions. Due to a lack of a more scalable inference algorithm, despite the usefulness, DTMs have not captured large topic dynamics. This paper fills this research void, and presents a fast and parallelizable inference algorithm using Gibbs Sampling with Stochastic Gradient Langevin Dynamics that does not make any unwarranted assumptions. We also present a Metropolis-Hastings based $O(1)$ sampler for topic assignments for each word token. In a distributed environment, our algorithm requires very little communication between workers during sampling (almost embarrassingly parallel) and scales up to large-scale applications. We are able to learn the largest Dynamic Topic Model to our knowledge, and learned the dynamics of 1,000 topics from 2.6 million documents in less than half an hour, and our empirical results show that our algorithm is not only orders of magnitude faster than the baselines but also achieves lower perplexity.

Developing efficient and scalable algorithms for Latent Dirichlet Allocation (LDA) is of wide interest for many applications. Previous work has developed an O (1) Metropolis-Hastings (MH) sampling method for each token. However, its performance is far from being optimal due to frequent cache misses caused by random accesses to the parameter matrices. In this paper, we first carefully analyze the memory access behavior of existing algorithms for LDA by cache locality at document level. We then develop WarpLDA, which achieves O (1) time complexity per-token and fits the randomly accessed memory perdocument in the L3 cache. Our empirical results in a wide range of testing conditions demonstrate that WarpLDA is consistently 5-15x faster than the state-of-the-art MH-based LightLDA, and is faster than the state-of-the-art sparsity aware F+LDA in most settings. Our WarpLDA learns a million topics from 639 millions of documents in only five hours at an unprecedented throughput of 11 billion tokens per second.

The rise of crowdsourcing brings new types of malpractices in Internet advertising. One can easily hire web workers through malicious crowdsourcing platforms to attack other advertisers. Such human generated crowd frauds are hard to detect by conventional fraud detection methods. In this paper, we carefully examine the characteristics of the group behaviors of crowd fraud and identify three persistent patterns, which are moderateness, synchronicity and dispersivity. Then we propose an effective crowd fraud detection method for search engine advertising based on these patterns, which consists of a constructing stage, a clustering stage and a filtering stage. At the constructing stage, we remove irrelevant data and reorganize the click logs into a surfer-advertiser inverted list; At the clustering stage, we define the sync-similarity between surfers' click histories and transform the coalition detection to a clustering problem, solved by a nonparametric algorithm; and finally we build a dispersity filter to remove false alarm clusters. The nonparametric nature of our method ensures that we can find an unbounded number of coalitions with nearly no human interaction. We also provide a parallel solution to make the method scalable to Web data and conduct extensive experiments. The empirical results demonstrate that our method is accurate and scalable.

Learning with multi-modal data is at the core of many multimedia applications, such as cross-modal retrieval and image annotation. In this paper, we present a nonparametric Bayesian approach to learning upstream supervised topic models for analyzing multi-modal data. Our model develops a compound nonparametric Bayesian multi-modal prior to describe the correlation structure of data both within each individual modality and between different modalities. It extends the hierarchical Dirichlet process (HDP) through incorporating upstream supervised response variables and values of latent functions under Gaussian process (GP). Upstream responses shared by data from multiple modalities are beneficial for discriminatively training and GP allows flexible structure learning of correlations. Hence, our model inherits the automatic determination of the number of topics from HDP, structure learning from GP and enhanced predictive capacity from upstream supervision. We also provide efficient variational inference and prediction algorithms. Empirical studies demonstrate superior performances on several benchmark datasets compared with previous competitors.

Topic models have played a pivotal role in analyzing large collections of complex data. Besides discovering latent semantics, supervised topic models (STMs) can make predictions on unseen test data. By marrying with advanced learning techniques, the predictive strengths of STMs have been dramatically enhanced, such as max-margin supervised topic models, state-of-the-art methods that integrate max-margin learning with topic models. Though powerful, max-margin STMs have a hard non-smooth learning problem. Existing algorithms rely on solving multiple latent SVM subproblems in an EM-type procedure, which can be too slow to be applicable to large-scale categorization tasks.

Topic models have shown great promise in discovering latent semantic structures from complex data corpora, ranging from text documents and web news articles to images, videos, and even biological data. In order to deal with massive data collections and dynamic text streams, probabilistic online topic models such as online latent Dirichlet allocation (OLDA) have recently been developed. However, due to normalization constraints, OLDA can be ineffective in controlling the sparsity of discovered representations, a desirable property for learning interpretable semantic patterns, especially when the total number of topics is large. In contrast, sparse topical coding (STC) has been successfully introduced as a non-probabilistic topic model for effectively discovering sparse latent patterns by using sparsity-inducing regularization. But, unfortunately STC cannot scale to very large datasets or deal with online text streams, partly due to its batch learning procedure. In this paper, we present a sparse online topic model, which directly controls the sparsity of latent semantic patterns by imposing sparsity-inducing regularization and learns the topical dictionary by an online algorithm. The online algorithm is efficient and guaranteed to converge. Extensive empirical results of the sparse online topic model as well as its collapsed and supervised extensions on a large-scale Wikipedia dataset and the medium-sized 20Newsgroups dataset demonstrate appealing performance.

Crowdsourcing has recently become popular among machine learning researchers and social scientists as an effective way to collect large-scale experimental data from distributed workers. To extract useful information from the cheap but potentially unreliable answers to tasks, a key problem is to identify reliable workers as well as unambiguous tasks. Although for objective tasks that have one correct answer per task, previous works can estimate worker reliability and task clarity based on the single gold standard assumption, for tasks that are subjective and accept multiple reasonable answers that workers may be grouped into, a phenomenon called schools of thought, existing models cannot be trivially applied. In this work, we present a statistical model to estimate worker reliability and task clarity without resorting to the single gold standard assumption. This is instantiated by explicitly characterizing the grouping behavior to form schools of thought with a rank-1 factorization of a worker-task groupsize matrix. Instead of performing an intermediate inference step, which can be expensive and unstable, we present an algorithm to analytically compute the sizes of different groups. We perform extensive empirical studies on real data collected from Amazon Mechanical Turk. Our method discovers the schools of thought, shows reasonable estimation of worker reliability and task clarity, and is robust to hyperparameter changes. Furthermore, our estimated worker reliability can be used to improve the gold standard prediction for objective tasks.

Probabilistic topic models have shown remarkable success in many application domains. However, a probabilistic conditional topic model can be extremely inefficient when considering a rich set of features because it needs to define a normalized distribution, which usually involves a hard-to-compute partition function. This paper presents conditional topical coding (CTC), a novel formulation of conditional topic models which is non-probabilistic. CTC relaxes the normalization constraints as in probabilistic models and learns non-negative document codes and word codes. CTC does not need to define a normalized distribution and can efficiently incorporate a rich set of features for improved topic discovery and prediction tasks. Moreover, CTC can directly control the sparsity of inferred representations by using appropriate regularization. We develop an efficient and easy-to-implement coordinate descent learning algorithm, of which each coding substep has a closed-form solution. Finally, we demonstrate the advantages of CTC on online review analysis datasets. Our results show that conditional topical coding can achieve state-of-the-art prediction performance and is much more efficient in training (one order of magnitude faster) and testing (two orders of magnitude faster) than probabilistic conditional topic models.

Feature selection is an important task in order to achieve better generalizability in high dimensional learning, and structure learning of Markov random fields (MRFs) can automatically discover the inherent structures underlying complex data. Both problems can be cast as solving an l1-norm regularized parameter estimation problem. The existing Grafting method can avoid doing inference on dense graphs in structure learning by incrementally selecting new features. However, Grafting performs a greedy step to optimize over free parameters once new features are included. This greedy strategy results in low efficiency when parameter learning is itself non-trivial, such as in MRFs, in which parameter learning depends on an expensive subroutine to calculate gradients. The complexity of calculating gradients in MRFs is typically exponential to the size of maximal cliques.

Online forums represent one type of social media that is particularly rich for studying human behavior in information seeking and diffusing. The way users join communities is a reflection of the changing and expanding of their interests toward information. In this paper, we study the patterns of user participation behavior, and the feature factors that influence such behavior on different forum datasets. We find that, despite the relative randomness and lesser commitment of structural relationships in online forums, users’ community joining behaviors display some strong regularities. One particularly interesting observation is that the very weak relationships between users defined by online replies have similar diffusion curves as those of real friendships or co-authorships. We build social selection models, Bipartite Markov Random Field (BiMRF), to quantitatively evaluate the prediction performance of those feature factors and their relationships. Using these models, we show that some features carry supplementary information, and the effectiveness of different features vary in different types of forums. Moreover, the results of BiMRF with two-star configurations suggest that the feature of user similarity defined by frequency of communication or number of common friends is inadequate to predict grouping behavior, but adding node-level features can improve the fit of the model.

Max-margin Markov networks (M3N) have shown great promise in structured prediction and relational learning. Due to the KKT conditions, the M3N enjoys dual sparsity. However, the existing M3N formulation does not enjoy primal sparsity, which is a desirable property for selecting significant features and reducing the risk of over-fitting. In this paper, we present an l1-norm regularized max-margin Markov network (l1-M3N), which enjoys dual and primal sparsity simultaneously. To learn an l1-M3N, we present three methods including projected sub-gradient, cutting-plane, and a novel EM-style algorithm, which is based on an equivalence between l1-M3N and an adaptive M3N. We perform extensive empirical studies on both synthetic and real data sets. Our experimental results show that: (1) l1-M3N can effectively select significant features; (2) l1-M3N can perform as well as the pseudo-primal sparse Laplace M3N in prediction accuracy, while consistently outperforms other competing methods that enjoy either primal or dual sparsity; and (3) the EM-algorithm is more robust than the other two in pre-diction accuracy and time efficiency.

Web forums have become an important data resource for many web applications, but extracting structured data from unstructured web forum pages is still a challenging task due to both complex page layout designs and unrestricted user created posts. In this paper, we study the problem of structured data extraction from various web forum sites. Our target is to find a solution as general as possible to extract structured data, such as post title, post author, post time, and post content from any forum site. In contrast to most existing information extraction methods, which only leverage the knowledge inside an individual page, we incorporate both page-level and site-level knowledge and employ Markov logic networks (MLNs) to effectively integrate all useful evidence by learning their importance automatically. Site-level knowledge includes (1) the linkages among different object pages, such as list pages and post pages, and (2) the interrelationships of pages belonging to the same object. The experimental results on 20 forums show a very encouraging information extraction performance, and demonstrate the ability of the proposed approach on various forums. We also show that the performance is limited if only page-level knowledge is used, while when incorporating the site-level knowledge both precision and recall can be significantly improved.

Traditional relation extraction methods require pre-specified relations and relation-specific human-tagged examples. Bootstrapping systems significantly reduce the number of training examples, but they usually apply heuristic-based methods to combine a set of strict hard rules, which limit the ability to generalize and thus generate a low recall. Furthermore, existing bootstrapping methods do not perform open information extraction (Open IE), which can identify various types of relations without requiring pre-specifications. In this paper, we propose a statistical extraction framework called Statistical Snowball (StatSnowball), which is a bootstrapping system and can perform both traditional relation extraction and Open IE.

Recent work has shown the effectiveness of leveraging layout and tag-tree structure for segmenting webpages and labeling HTML elements. However, how to effectively segment and label the text contents inside HTML elements is still an open problem. Since many text contents on a webpage are often text fragments and not strictly grammatical, traditional natural language processing techniques, that typically expect grammatical sentences, are no longer directly applicable. In this paper, we examine how to use layout and tag-tree structure in a principled way to help understand text contents on webpages. We propose to segment and label the page structure and the text content of a webpage in a joint discriminative probabilistic model. In this model, semantic labels of page structure can be leveraged to help text content understanding, and semantic labels ofthe text phrases can be used in page structure understanding tasks such as data record detection. Thus, integration of both page structure and text content understanding leads to an integrated solution of webpage understanding. Experimental results on research homepage extraction show the feasibility and promise of our approach.

Recent work has shown the feasibility and promise of templateindependent Web data extraction. However, existing approaches use decoupled strategies – attempting to do data record detection and attribute labeling in two separate phases. In this paper, we show that separately extracting data records and attributes is highly ineffective and propose a probabilistic model to perform these two tasks simultaneously. In our approach, record detection can benefit from the availability of semantics required in attribute labeling and, at the same time, the accuracy of attribute labeling can be improved when data records are labeled in a collective manner. The proposed model is called Hierarchical Conditional Random Fields. It can efficiently integrate all useful features by learning their importance, and it can also incorporate hierarchical interactions which are very important for Web data extraction. We empirically compare the proposed model with existing decoupled approaches for product information extraction, and the results show significant improvements in both record detection and attribute labeling.