We propose Social Diffusion, a novel method for short-term and long-term forecasting of the motion of multiple persons as well as their social interactions. Jointly forecasting motions for multiple persons involved in social activities is inherently a challenging problem due to the interdependencies between individuals. In this work, we leverage a diffusion model conditioned on motion histories and causal temporal convolutional networks to forecast individually and contextually plausible motions for all participants. The contextual plausibility is achieved via an order-invariant aggregation function. As a second contribution, we design a new evaluation protocol that measures the plausibility of social interactions which we evaluate on the Haggling dataset, which features a challenging social activity where people are actively taking turns to talk and switching their attention. We evaluate our approach on four datasets for multi-person forecasting where our approach outperforms the state-of-the-art in terms of motion realism and contextual plausibility.

Iterative Hard Thresholding (IHT) is a popular class of first-order greedy selection methods for loss minimization under cardinality constraint. The existing IHT-style algorithms, however, are proposed for minimizing the primal formulation. It is still an open issue to explore duality theory and algorithms for such a non-convex and NP-hard combinatorial optimization problem. To address this issue, we develop in this article a novel duality theory for $\ell_2$-regularized empirical risk minimization under cardinality constraint, along with an IHT-style algorithm for dual optimization. Our sparse duality theory establishes a set of sufficient and/or necessary conditions under which the original non-convex problem can be equivalently or approximately solved in a concave dual formulation. In view of this theory, we propose the Dual IHT (DIHT) algorithm as a super-gradient ascent method to solve the non-smooth dual problem with provable guarantees on primal-dual gap convergence and sparsity recovery. Numerical results confirm our theoretical predictions and demonstrate the superiority of DIHT to the state-of-the-art primal IHT-style algorithms in model estimation accuracy and computational efficiency.

In this paper, we present a sample distributed greedy pursuit method for non-convex sparse learning under cardinality constraint. Given the training samples uniformly randomly partitioned across multiple machines, the proposed method alternates between local inexact sparse minimization of a Newton-type approximation and centralized global results aggregation. Theoretical analysis shows that for a general class of convex functions with Lipschitze continues Hessian, the method converges linearly with contraction factor scaling inversely to the local data size; whilst the communication complexity required to reach desirable statistical accuracy scales logarithmically with respect to the number of machines for some popular statistical learning models. For nonconvex objective functions, up to a local estimation error, our method can be shown to converge to a local stationary sparse solution with sub-linear communication complexity. Numerical results demonstrate the efficiency and accuracy of our method when applied to large-scale sparse learning tasks including deep neural nets pruning

Recent developments in gradient-based attention modeling have seen attention maps emerge as a powerful tool for interpreting convolutional neural networks. Despite good localization for an individual class of interest, these techniques produce attention maps with substantially overlapping responses among different classes, leading to the problem of visual confusion and the need for discriminative attention. In this paper, we address this problem by means of a new framework that makes class-discriminative attention a principled part of the learning process. Our key innovations include new learning objectives for attention separability and cross-layer consistency, which result in improved attention discriminability and reduced visual confusion. Extensive experiments on image classification benchmarks show the effectiveness of our approach in terms of improved classification accuracy, including CIFAR-100 (+3.33%), Caltech-256 (+1.64%), ImageNet (+0.92%), CUB-200-2011 (+4.8%) and PASCAL VOC2012 (+5.73%).

We propose an end-to-end network for visual illustration of a sequence of sentences forming a story. At the core of our model is the ability to model the inter-related nature of the sentences within a story, as well as the ability to learn coherence to support reference resolution. The framework takes the form of an encoder-decoder architecture, where sentences are encoded using a hierarchical two-level sentence-story GRU, combined with an encoding of coherence, and sequentially decoded using a predicted feature representation into a consistent illustrative image sequence. We optimize all parameters of our network in an end-to-end fashion with respect to order embedding loss, encoding entailment between images and sentences. Experiments on the VIST storytelling dataset [9] highlight the importance of our algorithmic choices and efficacy of our overall model.

Face detection and landmark localization have been extensively investigated and are the prerequisite for many face applications, such as face recognition and 3D face reconstruction. Most existing methods achieve success on only one of the two problems. In this paper, we propose a coupled encoder-decoder network to jointly detect faces and localize facial key points. The encoder and decoder generate response maps for facial landmark localization. Moreover, we observe that the intermediate feature maps from the encoder and decoder have strong power in describing facial regions, which motivates us to build a unified framework by coupling the feature maps for multi-scale cascaded face detection. Experiments on face detection show strongly competitive results against the existing methods on two public benchmarks. The landmark localization further shows consistently better accuracy than state-of-the-arts on three face-in-the-wild databases.

Iterative Hard Thresholding (IHT) is a class of projected gradient descent methods for optimizing sparsity-constrained minimization models, with the best known efficiency and scalability in practice. As far as we know, the existing IHT-style methods are designed for sparse minimization in primal form. It remains open to explore duality theory and algorithms in such a non-convex and NP-hard setting. In this article, we bridge the gap by establishing a duality theory for sparsity-constrained minimization with $\ell_2$-regularized objective and proposing an IHT-style algorithm for dual maximization. Our sparse duality theory provides a set of sufficient and necessary conditions under which the original NP-hard/non-convex problem can be equivalently solved in a dual space. The proposed dual IHT algorithm is a super-gradient method for maximizing the non-smooth dual objective. An interesting finding is that the sparse recovery performance of dual IHT is invariant to the Restricted Isometry Property (RIP), which is required by all the existing primal IHT without sparsity relaxation. Moreover, a stochastic variant of dual IHT is proposed for large-scale stochastic optimization. Numerical results demonstrate that dual IHT algorithms can achieve more accurate model estimation given small number of training data and have higher computational efficiency than the state-of-the-art primal IHT-style algorithms.

The state of the art in query expansion is mainly based on the spatial information. These methods achieve high performance, however, suffer from huge computation and memory. The objective of this paper is to perform visual reranking in near-real time regardless of the spatial information. We explore a graph-based method proposed as our confident sample detection baseline, which has been proved successful in achieving high precision. In addition, a novel maximum-kernel-based metric function is introduced to rerank the images in the initial result. We evaluated the method on the standard Paris dataset and a new Francelandmark dataset. Our experiments demonstrate that the algorithm has great value on practicality because of its good performance, easy implementation, and high computational efficiency.

Image search reranking has become a widely-used approach to significantly boost retrieval performance in the state-of-art content-based image retrieval system. Most of the methods merely rely on matching visual distances between query and initial results or among initial results to detect confident samples relevant to query. However, they may fail to rerank due to the existence of a huge gap between low-level visual features and high-level semantic concepts. In this paper, we propose to detect reliable relevant samples based on a semantic image graph of labeled auxiliary dataset and Markov random walk algorithm. A graph-based rerank method is then presented to propagate the scores of detected confident samples to the rest. Our method is evaluated on the standard Paris dataset and a France dataset introduced by us. The performance is demonstrated to match or exceed the state-of-art.

The exponential growth of web videos brings content based copy detection into a crucial issue. Besides the image information, audio also plays an important role in copy detection. In this paper, the audio-based copy detection framework is introduced. Three contributions are presented: (1) the band energy difference based feature is improved by adding multi-scale information, which extends the candidate feature sets, (2) a conditional entropy based method is used to select 16 ordinal relations to generate a more compact and robust feature combination among the random $C_{91}^{16}\approx2.6\times10^{17}$ combinations, (3) the result-based fusion strategy is introduced to recall the missed true positives. The proposed algorithm outperforms the traditional coarse fingerprints, shown by experiments conducted in the TRECVID 2011 Content-based Copy Detection (CCD) database.

A TV program segmentation algorithm is presented by the fusion of the multi-modal information in the large-scale videos. As "Inter-Programs" are generally inserted into the TV videos repeatedly, the macro structures of the videos can be effectively and automatically generated by identifying the video-audio features of the special sequences. The Electronic Program Guide (EPG) is used to organize the structures into the programs. Three sections are included in the algorithm, namely, the video-based non-supervised duplicate sequence detection, the audio-based special clip retrieval and the EPG-based 24-hour program segmentation. The algorithm has been tested in 60-day different-type TV videos. The F-measures of the multi-modal fusion and video-based duplicated sequence detection achieve the rates of over 98% and 96% respectively. These results show that the proposed method is highly efficient and effective for the TV Program segmentation.