We study the problem of community detection in multi-layer networks, where pairs of nodes can be related in multiple modalities. We introduce a general framework, i.e., mixture multi-layer stochastic block model (MMSBM), which includes many earlier models as special cases. We propose a tensor-based algorithm (TWIST) to reveal both global/local memberships of nodes, and memberships of layers. We show that the TWIST procedure can accurately detect the communities with small misclassification error as the number of nodes and/or the number of layers increases. Numerical studies confirm our theoretical findings. To our best knowledge, this is the first systematic study on the mixture multi-layer networks using tensor decomposition. The method is applied to two real datasets: worldwide trading networks and malaria parasite genes networks, yielding new and interesting findings.
We study the design of stochastic local search methods to prove unsatisfiability of a constraint satisfaction problem (CSP). For a binary CSP, such methods have been designed using the microstructure of the CSP. Here, we develop a method to decompose the microstructure into graph tensors. We show how to use the tensor decomposition to compute a proof of unsatisfiability efficiently and in parallel. We also offer substantial empirical evidence that our approach improves the praxis. For instance, one decomposition yields proofs of unsatisfiability in half the time without sacrificing the quality. Another decomposition is twenty times faster and effective three-tenths of the times compared to the prior method. Our method is applicable to arbitrary CSPs using the well known dual and hidden variable transformations from an arbitrary CSP to a binary CSP.
Coupled tensor decomposition reveals the joint data structure by incorporating priori knowledge that come from the latent coupled factors. The tensor ring (TR) decomposition is invariant under the permutation of tensors with different mode properties, which ensures the uniformity of decomposed factors and mode attributes. The TR has powerful expression ability and achieves success in some multi-dimensional data processing applications. To let coupled tensors help each other for missing component estimation, in this paper we utilize TR for coupled completion by sharing parts of the latent factors. The optimization model for coupled TR completion is developed with a novel Frobenius norm. It is solved by the block coordinate descent algorithm which efficiently solves a series of quadratic problems resulted from sampling pattern. The excess risk bound for this optimization model shows the theoretical performance enhancement in comparison with other coupled nuclear norm based methods. The proposed method is validated on numerical experiments on synthetic data, and experimental results on real-world data demonstrate its superiority over the state-of-the-art methods in terms of recovery accuracy.
There is currently an unprecedented demand for large-scale temporal data analysis due to the explosive growth of data. Dynamic topic modeling has been widely used in social and data sciences with the goal of learning latent topics that emerge, evolve, and fade over time. Previous work on dynamic topic modeling primarily employ the method of nonnegative matrix factorization (NMF), where slices of the data tensor are each factorized into the product of lower-dimensional nonnegative matrices. With this approach, however, information contained in the temporal dimension of the data is often neglected or underutilized. To overcome this issue, we propose instead adopting the method of nonnegative CANDECOMP/PARAPAC (CP) tensor decomposition (NNCPD), where the data tensor is directly decomposed into a minimal sum of outer products of nonnegative vectors, thereby preserving the temporal information. The viability of NNCPD is demonstrated through application to both synthetic and real data, where significantly improved results are obtained compared to those of typical NMF-based methods. The advantages of NNCPD over such approaches are studied and discussed. To the best of our knowledge, this is the first time that NNCPD has been utilized for the purpose of dynamic topic modeling, and our findings will be transformative for both applications and further developments.
Data collected at very frequent intervals is usually extremely sparse and has no structure that is exploitable by modern tensor decomposition algorithms. Thus the utility of such tensors is low, in terms of the amount of interpretable and exploitable structure that one can extract from them. In this paper, we introduce the problem of finding a tensor of adaptive aggregated granularity that can be decomposed to reveal meaningful latent concepts (structures) from datasets that, in their original form, are not amenable to tensor analysis. Such datasets fall under the broad category of sparse point processes that evolve over space and/or time. To the best of our knowledge, this is the first work that explores adaptive granularity aggregation in tensors. Furthermore, we formally define the problem and discuss what different definitions of "good structure" can be in practice, and show that optimal solution is of prohibitive combinatorial complexity. Subsequently, we propose an efficient and effective greedy algorithm called IceBreaker, which follows a number of intuitive decision criteria that locally maximize the "goodness of structure", resulting in high-quality tensors. We evaluate our method on synthetic, semi-synthetic and real datasets. In all the cases, our proposed method constructs tensors that have very high structure quality.
Low-rank tensor decomposition and completion have attracted significant interest from academia given the ubiquity of tensor data. However, the low-rank structure is a global property, which will not be fulfilled when the data presents complex and weak dependencies given specific graph structures. One particular application that motivates this study is the spatiotemporal data analysis. As shown in the preliminary study, weakly dependencies can worsen the low-rank tensor completion performance. In this paper, we propose a novel low-rank CANDECOMP / PARAFAC (CP) tensor decomposition and completion framework by introducing the $L_{1}$-norm penalty and Graph Laplacian penalty to model the weakly dependency on graph. We further propose an efficient optimization algorithm based on the Block Coordinate Descent for efficient estimation. A case study based on the metro passenger flow data in Hong Kong is conducted to demonstrate improved performance over the regular tensor completion methods.
Speech-driven facial animation involves using a speech signal to generate realistic videos of talking faces. Recent deep learning approaches to facial synthesis rely on extracting low-dimensional representations and concatenating them, followed by a decoding step of the concatenated vector. This accounts for only first-order interactions of the features and ignores higher-order interactions. In this paper we propose a polynomial fusion layer that models the joint representation of the encodings by a higher-order polynomial, with the parameters modelled by a tensor decomposition. We demonstrate the suitability of this approach through experiments on generated videos evaluated on a range of metrics on video quality, audiovisual synchronisation and generation of blinks.
Portfolio allocation and risk management make use of correlation matrices and heavily rely on the choice of a proper correlation matrix to be used. In this regard, one important question is related to the choice of the proper sample period to be used to estimate a stable correlation matrix. This paper addresses this question and proposes a new methodology to estimate the correlation matrix which doesn't depend on the chosen sample period. This new methodology is based on tensor factorization techniques. In particular, combining and normalizing factor components, we build a correlation matrix which shows emerging structural dependency properties not affected by the sample period. To retrieve the factor components, we propose a new tensor decomposition (which we name Slice-Diagonal Tensor (SDT) factorization) and compare it to the two most used tensor decompositions, the Tucker and the PARAFAC. We have that the new factorization is more parsimonious than the Tucker decomposition and more flexible than the PARAFAC. Moreover, this methodology applied to both simulated and empirical data shows results which are robust to two non-parametric tests, namely Kruskal-Wallis and Kolmogorov-Smirnov tests. Since the resulting correlation matrix features stability and emerging structural dependency properties, it can be used as alternative to other correlation matrices type of measures, including the Person correlation.
We analyze large, multi-dimensional, sparse counting data sets, finding unsupervised groups to provide unique insights into genetic data. We create gene and biological pathway groups based on patients' variants to find common risk factors for four common types of cancer (breast, lung, prostate, and colorectal) and autism spectrum disorder. To accomplish this, we extend latent Dirichlet allocation to multiple dimensions and design distinct methods for hierarchical topic modeling. We find that our conditional hierarchical Bayesian Tucker decomposition models are more coherent than baseline models.
We study the optimization problem for decomposing $d$ dimensional fourth-order Tensors with $k$ non-orthogonal components. We derive \textit{deterministic} conditions under which such a problem does not have spurious local minima. In particular, we show that if $\kappa = \frac{\lambda_{max}}{\lambda_{min}} < \frac{5}{4}$, and incoherence coefficient is of the order $O(\frac{1}{\sqrt{d}})$, then all the local minima are globally optimal. Using standard techniques, these conditions could be easily transformed into conditions that would hold with high probability in high dimensions when the components are generated randomly. Finally, we prove that the tensor power method with deflation and restarts could efficiently extract all the components within a tolerance level $O(\kappa \sqrt{k\tau^3})$ that seems to be the noise floor of non-orthogonal tensor decomposition.
We address the problem of Visual Relationship Detection (VRD) which aims to describe the relationships between pairs of objects in the form of triplets of (subject, predicate, object). We observe that given a pair of bounding box proposals, objects often participate in multiple relations implying the distribution of triplets is multimodal. We leverage the strong correlations within triplets to learn the joint distribution of triplet variables conditioned on the image and the bounding box proposals, doing away with the hitherto used independent distribution of triplets. To make learning the triplet joint distribution feasible, we introduce a novel technique of learning conditional triplet distributions in the form of their normalized low rank non-negative tensor decompositions. Normalized tensor decompositions take form of mixture distributions of discrete variables and thus are able to capture multimodality. This allows us to efficiently learn higher order discrete multimodal distributions and at the same time keep the parameter size manageable. We further model the probability of selecting an object proposal pair and include a relation triplet prior in our model. We show that each part of the model improves performance and the combination outperforms state-of-the-art score on the Visual Genome (VG) and Visual Relationship Detection (VRD) datasets.
MicroRNAs (miRNAs) play crucial roles in multifarious biological processes associated with human diseases. Identifying potential miRNA-disease associations contributes to understanding the molecular mechanisms of miRNA-related diseases. Most of the existing computational methods mainly focus on predicting whether a miRNA-disease association exists or not. However, the roles of miRNAs in diseases are prominently diverged, for instance, Genetic variants of microRNA (mir-15) may affect expression level of miRNAs leading to B cell chronic lymphocytic leukemia, while circulating miRNAs (including mir-1246, mir-1307-3p, etc.) have potentials to detecting breast cancer in the early stage. In this paper, we aim to predict multi-type miRNA-disease associations instead of taking them as binary. To this end, we innovatively represent miRNA-disease-type triplets as a tensor and introduce Tensor Decomposition methods to solve the prediction task. Experimental results on two widely-adopted miRNA-disease datasets: HMDD v2.0 and HMDD v3.2 show that tensor decomposition methods improve a recent baseline in a large scale (up to 38% in top-1 F1). We further propose a novel method, Tensor Decomposition with Relational Constraints (TDRC), which incorporates biological features as relational constraints to further the existing tensor decomposition methods. Compared with two existing tensor decomposition methods, TDRC can produce better performance while being more efficient.
Visual objects are composed of a recursive hierarchy of perceptual wholes and parts, whose properties, such as shape, reflectance, and color, constitute a hierarchy of intrinsic causal factors of object appearance. However, object appearance is the compositional consequence of both an object's intrinsic and extrinsic causal factors, where the extrinsic causal factors are related to illumination, and imaging conditions. Therefore, this paper proposes a unified tensor model of wholes and parts, and introduces a compositional hierarchical tensor factorization that disentangles the hierarchical causal structure of object image formation, and subsumes multilinear block tensor decomposition as a special case. The resulting object representation is an interpretable combinatorial choice of wholes' and parts' representations that renders object recognition robust to occlusion and reduces training data requirements. We demonstrate ourapproach in the context of face recognition by training on an extremely reduced dataset of synthetic images, and report encouragingface verification results on two datasets - the Freiburg dataset, andthe Labeled Face in the Wild (LFW) dataset consisting of real world images, thus, substantiating the suitability of our approach for data starved domains.
We present Nesterov-type acceleration techniques for Alternating Least Squares (ALS) methods applied to canonical tensor decomposition. While Nesterov acceleration turns gradient descent into an optimal first-order method for convex problems by adding a momentum term with a specific weight sequence, a direct application of this method and weight sequence to ALS results in erratic convergence behaviour. This is so because the tensor decomposition problem is non-convex and ALS is accelerated instead of gradient descent. Instead, we consider various restart mechanisms and suitable choices of momentum weights that enable effective acceleration. Our extensive empirical results show that the Nesterov-accelerated ALS methods with restart can be dramatically more efficient than the stand-alone ALS or Nesterov accelerated gradient methods, when problems are ill-conditioned or accurate solutions are desired. The resulting methods perform competitively with or superior to existing acceleration methods for ALS, including ALS acceleration by NCG, NGMRES, or LBFGS, and additionally enjoy the benefit of being much easier to implement. We also compare with Nesterov-type updates where the momentum weight is determined by a line search, which are equivalent or closely related to existing line search methods for ALS. On a large and ill-conditioned 71$\times$1000$\times$900 tensor consisting of readings from chemical sensors to track hazardous gases, the restarted Nesterov-ALS method shows desirable robustness properties and outperforms any of the existing methods by a large factor. There is clear potential for extending our Nesterov-type acceleration approach to accelerating other optimization algorithms than ALS applied to other non-convex problems, such as Tucker tensor decomposition. Our Matlab code is available at https://github.com/hansdesterck/nonlinear-preconditioning-for-optimization.
This paper addresses the challenging task of video captioning which aims to generate descriptions for video data. Recently, the attention-based encoder-decoder structures have been widely used in video captioning. In existing literature, the attention weights are often built from the information of an individual modality, while, the association relationships between multiple modalities are neglected. Motivated by this observation, we propose a video captioning model with High-Order Cross-Modal Attention (HOCA) where the attention weights are calculated based on the high-order correlation tensor to capture the frame-level cross-modal interaction of different modalities sufficiently. Furthermore, we novelly introduce Low-Rank HOCA which adopts tensor decomposition to reduce the extremely large space requirement of HOCA, leading to a practical and efficient implementation in real-world applications. Experimental results on two benchmark datasets, MSVD and MSR-VTT, show that Low-rank HOCA establishes a new state-of-the-art.
Higher-order tensors have received increased attention across science and engineering. While most tensor decomposition methods are developed for a single tensor observation, scientific studies often collect side information, in the form of node features and interactions thereof, together with the tensor data. Such data problems are common in neuroimaging, network analysis, and spatial-temporal modeling. Identifying the relationship between a high-dimensional tensor and side information is important yet challenging. Here, we develop a tensor decomposition method that incorporates multiple feature matrices as side information. Unlike unsupervised tensor decomposition, our supervised decomposition captures the effective dimension reduction of the data tensor confined to feature space of interest. An efficient alternating optimization algorithm with provable spectral initialization is further developed. Our proposal handles a broad range of data types, including continuous, count, and binary observations. We apply the method to diffusion tensor imaging data from human connectome project and multi-relational political network data. We identify the key global connectivity pattern and pinpoint the local regions that are associated with available features. Our simulation code, R-package tensorregress, and datasets used in the paper are available at https://CRAN.R-project.org/package=tensorregress.
How can we efficiently compress Convolutional Neural Network (CNN) while retaining their accuracy on classification tasks? Depthwise Separable Convolution (DSConv), which replaces a standard convolution with a depthwise convolution and a pointwise convolution, has been used for building lightweight architectures. However, previous works based on depthwise separable convolution are limited when compressing a trained CNN model since 1) they are mostly heuristic approaches without a precise understanding of their relations to standard convolution, and 2) their accuracies do not match that of the standard convolution. In this paper, we propose FALCON, an accurate and lightweight method to compress CNN. FALCON uses GEP, our proposed mathematical formulation to approximate the standard convolution kernel, to interpret existing convolution methods based on depthwise separable convolution. By exploiting the knowledge of a trained standard model and carefully determining the order of depthwise separable convolution via GEP, FALCON achieves sufficient accuracy close to that of the trained standard model. Furthermore, this interpretation leads to developing a generalized version rank-k FALCON which performs k independent FALCON operations and sums up the result. Experiments show that FALCON 1) provides higher accuracy than existing methods based on depthwise separable convolution and tensor decomposition, and 2) reduces the number of parameters and FLOPs of standard convolution by up to a factor of 8 while ensuring similar accuracy. We also demonstrate that rank-k FALCON further improves the accuracy while sacrificing a bit of compression and computation reduction rates.
Tensor decomposition is one of the well-known approaches to reduce the latency time and number of parameters of a pre-trained model. However, in this paper, we propose an approach to use tensor decomposition to reduce training time of training a model from scratch. In our approach, we train the model from scratch (i.e., randomly initialized weights) with its original architecture for a small number of epochs, then the model is decomposed, and then continue training the decomposed model till the end. There is an optional step in our approach to convert the decomposed architecture back to the original architecture. We present results of using this approach on both CIFAR10 and Imagenet datasets, and show that there can be upto 2x speed up in training time with accuracy drop of upto 1.5% only, and in other cases no accuracy drop. This training acceleration approach is independent of hardware and is expected to have similar speed ups on both CPU and GPU platforms.
Event detection is gaining increasing attention in smart cities research. Large-scale mobility data serves as an important tool to uncover the dynamics of urban transportation systems, and more often than not the dataset is incomplete. In this article, we develop a method to detect extreme events in large traffic datasets, and to impute missing data during regular conditions. Specifically, we propose a robust tensor recovery problem to recover low rank tensors under fiber-sparse corruptions with partial observations, and use it to identify events, and impute missing data under typical conditions. Our approach is scalable to large urban areas, taking full advantage of the spatio-temporal correlations in traffic patterns. We develop an efficient algorithm to solve the tensor recovery problem based on the alternating direction method of multipliers (ADMM) framework. Compared with existing $l_1$ norm regularized tensor decomposition methods, our algorithm can exactly recover the values of uncorrupted fibers of a low rank tensor and find the positions of corrupted fibers under mild conditions. Numerical experiments illustrate that our algorithm can exactly detect outliers even with missing data rates as high as 40%, conditioned on the outlier corruption rate and the Tucker rank of the low rank tensor. Finally, we apply our method on a real traffic dataset corresponding to downtown Nashville, TN, USA and successfully detect the events like severe car crashes, construction lane closures, and other large events that cause significant traffic disruptions.
Tensor decomposition methods are widely used for model compression and fast inference in convolutional neural networks (CNNs). Although many decompositions are conceivable, only CP decomposition and a few others have been applied in practice, and no extensive comparisons have been made between available methods. Previous studies have not determined how many decompositions are available, nor which of them is optimal. In this study, we first characterize a decomposition class specific to CNNs by adopting a flexible graphical notation. The class includes such well-known CNN modules as depthwise separable convolution layers and bottleneck layers, but also previously unknown modules with nonlinear activations. We also experimentally compare the tradeoff between prediction accuracy and time/space complexity for modules found by enumerating all possible decompositions, or by using a neural architecture search. We find some nonlinear decompositions outperform existing ones.