Ordinal regression aims to classify instances into ordinal categories. In this paper, body mass index (BMI) category estimation from facial images is cast as an ordinal regression problem. In particular, noisy binary search algorithms based on pairwise comparisons are employed to exploit the ordinal relationship among BMI categories. Comparisons are performed with Siamese architectures, one of which uses the Bradley-Terry model probabilities as target. The Bradley-Terry model is an approach to describe probabilities of the possible outcomes when elements of a set are repeatedly compared with one another in pairs. Experimental results show that our approach outperforms classification and regression-based methods at estimating BMI categories.
https://arxiv.org/abs/1811.03268
We study the problem of learning conditional generators from noisy labeled samples, where the labels are corrupted by random noise. A standard training of conditional GANs will not only produce samples with wrong labels, but also generate poor quality samples. We consider two scenarios, depending on whether the noise model is known or not. When the distribution of the noise is known, we introduce a novel architecture which we call Robust Conditional GAN (RCGAN). The main idea is to corrupt the label of the generated sample before feeding to the adversarial discriminator, forcing the generator to produce samples with clean labels. This approach of passing through a matching noisy channel is justified by corresponding multiplicative approximation bounds between the loss of the RCGAN and the distance between the clean real distribution and the generator distribution. This shows that the proposed approach is robust, when used with a carefully chosen discriminator architecture, known as projection discriminator. When the distribution of the noise is not known, we provide an extension of our architecture, which we call RCGAN-U, that learns the noise model simultaneously while training the generator. We show experimentally on MNIST and CIFAR-10 datasets that both the approaches consistently improve upon baseline approaches, and RCGAN-U closely matches the performance of RCGAN.
https://arxiv.org/abs/1811.03205
Recurrent Neural Networks (RNNs) are an important class of neural networks designed to retain and incorporate context into current decisions. RNNs are particularly well suited for machine learning problems in which context is important, such as speech recognition or language translation. This work presents RNNFast, a hardware accelerator for RNNs that leverages an emerging class of non-volatile memory called domain-wall memory (DWM). We show that DWM is very well suited for RNN acceleration due to its very high density and low read/write energy. At the same time, the sequential nature of input/weight processing of RNNs mitigates one of the downsides of DWM, which is the linear (rather than constant) data access time. RNNFast is very efficient and highly scalable, with flexible mapping of logical neurons to RNN hardware blocks. The basic hardware primitive, the RNN processing element (PE) includes custom DWM-based multiplication, sigmoid and tanh units for high density and low-energy. The accelerator is designed to minimize data movement by closely interleaving DWM storage and computation. We compare our design with a state-of-the-art GPGPU and find 21.8x better performance with 70x lower energy.
https://arxiv.org/abs/1812.07609
Data selection techniques applied to neural machine translation (NMT) aim to increase the performance of a model by retrieving a subset of sentences for use as training data. One of the possible data selection techniques are transductive learning methods, which select the data based on the test set, i.e. the document to be translated. A limitation of these methods to date is that using the source-side test set does not by itself guarantee that sentences are selected with correct translations, or translations that are suitable given the test-set domain. Some corpora, such as subtitle corpora, may contain parallel sentences with inaccurate translations caused by localization or length restrictions. In order to try to fix this problem, in this paper we propose to use an approximated target-side in addition to the source-side when selecting suitable sentence-pairs for training a model. This approximated target-side is built by pre-translating the source-side. In this work, we explore the performance of this general idea for one specific data selection approach called Feature Decay Algorithms (FDA). We train German-English NMT models on data selected by using the test set (source), the approximated target side, and a mixture of both. Our findings reveal that models built using a combination of outputs of FDA (using the test set and an approximated target side) perform better than those solely using the test set. We obtain a statistically significant improvement of more than 1.5 BLEU points over a model trained with all data, and more than 0.5 BLEU points over a strong FDA baseline that uses source-side information only.
https://arxiv.org/abs/1811.03039
The detection and the quantification of anomalies in image data are critical tasks in industrial scenes such as detecting micro scratches on product. In recent years, due to the difficulty of defining anomalies and the limit of correcting their labels, research on unsupervised anomaly detection using generative models has attracted attention. Generally, in those studies, only normal images are used for training to model the distribution of normal images. The model measures the anomalies in the target images by reproducing the most similar images and scoring image patches indicating their fit to the learned distribution. This approach is based on a strong presumption; the trained model should not be able to generate abnormal images. However, in reality, the model can generate abnormal images mainly due to noisy normal data which include small abnormal pixels, and such noise severely affects the accuracy of the model. Therefore, we propose a novel anomaly detection method to distort the distribution of the model with existing abnormal images. The proposed method detects pixel-level micro anomalies with a high accuracy from 1024x1024 high resolution images which are actually used in an industrial scene. In this paper, we share experimental results on open datasets, due to the confidentiality of the data.
https://arxiv.org/abs/1807.01136
In this paper, we propose a novel regularization method for Generative Adversarial Networks, which allows the model to learn discriminative yet compact binary representations of image patches (image descriptors). We employ the dimensionality reduction that takes place in the intermediate layers of the discriminator network and train binarized low-dimensional representation of the penultimate layer to mimic the distribution of the higher-dimensional preceding layers. To achieve this, we introduce two loss terms that aim at: (i) reducing the correlation between the dimensions of the binarized low-dimensional representation of the penultimate layer i. e. maximizing joint entropy) and (ii) propagating the relations between the dimensions in the high-dimensional space to the low-dimensional space. We evaluate the resulting binary image descriptors on two challenging applications, image matching and retrieval, and achieve state-of-the-art results.
https://arxiv.org/abs/1806.06778
Current image captioning approaches generate descriptions which lack specific information, such as named entities that are involved in the images. In this paper we propose a new task which aims to generate informative image captions, given images and hashtags as input. We propose a simple but effective approach to tackle this problem. We first train a convolutional neural networks - long short term memory networks (CNN-LSTM) model to generate a template caption based on the input image. Then we use a knowledge graph based collective inference algorithm to fill in the template with specific named entities retrieved via the hashtags. Experiments on a new benchmark dataset collected from Flickr show that our model generates news-style image descriptions with much richer information. Our model outperforms unimodal baselines significantly with various evaluation metrics.
https://arxiv.org/abs/1804.07889
A recent method employs 3D voxels to represent 3D shapes, but this limits the approach to low resolutions due to the computational cost caused by the cubic complexity of 3D voxels. Hence the method suffers from a lack of detailed geometry. To resolve this issue, we propose Y^2Seq2Seq, a view-based model, to learn cross-modal representations by joint reconstruction and prediction of view and word sequences. Specifically, the network architecture of Y^2Seq2Seq bridges the semantic meaning embedded in the two modalities by two coupled `Y’ like sequence-to-sequence (Seq2Seq) structures. In addition, our novel hierarchical constraints further increase the discriminability of the cross-modal representations by employing more detailed discriminative information. Experimental results on cross-modal retrieval and 3D shape captioning show that Y^2Seq2Seq outperforms the state-of-the-art methods.
https://arxiv.org/abs/1811.02745
In this paper we present a novel unsupervised representation learning approach for 3D shapes, which is an important research challenge as it avoids the manual effort required for collecting supervised data. Our method trains an RNN-based neural network architecture to solve multiple view inter-prediction tasks for each shape. Given several nearby views of a shape, we define view inter-prediction as the task of predicting the center view between the input views, and reconstructing the input views in a low-level feature space. The key idea of our approach is to implement the shape representation as a shape-specific global memory that is shared between all local view inter-predictions for each shape. Intuitively, this memory enables the system to aggregate information that is useful to better solve the view inter-prediction tasks for each shape, and to leverage the memory as a view-independent shape representation. Our approach obtains the best results using a combination of L_2 and adversarial losses for the view inter-prediction task. We show that VIP-GAN outperforms state-of-the-art methods in unsupervised 3D feature learning on three large scale 3D shape benchmarks.
https://arxiv.org/abs/1811.02744
In this paper, we propose an actor ensemble algorithm, named ACE, for continuous control with a deterministic policy in reinforcement learning. In ACE, we use actor ensemble (i.e., multiple actors) to search the global maxima of the critic. Besides the ensemble perspective, we also formulate ACE in the option framework by extending the option-critic architecture with deterministic intra-option policies, revealing a relationship between ensemble and options. Furthermore, we perform a look-ahead tree search with those actors and a learned value prediction model, resulting in a refined value estimation. We demonstrate a significant performance boost of ACE over DDPG and its variants in challenging physical robot simulators.
https://arxiv.org/abs/1811.02696
Recurrent neural networks (RNN) are at the core of modern automatic speech recognition (ASR) systems. In particular, long-short term memory (LSTM) recurrent neural networks have achieved state-of-the-art results in many speech recognition tasks, due to their efficient representation of long and short term dependencies in sequences of inter-dependent features. Nonetheless, internal dependencies within the element composing multidimensional features are weakly considered by traditional real-valued representations. We propose a novel quaternion long-short term memory (QLSTM) recurrent neural network that takes into account both the external relations between the features composing a sequence, and these internal latent structural dependencies with the quaternion algebra. QLSTMs are compared to LSTMs during a memory copy-task and a realistic application of speech recognition on the Wall Street Journal (WSJ) dataset. QLSTM reaches better performances during the two experiments with up to $2.8$ times less learning parameters, leading to a more expressive representation of the information.
https://arxiv.org/abs/1811.02566
The web contains vast repositories of unstructured text. We investigate the opportunity for building a knowledge graph from these text sources. We generate a set of triples which can be used in knowledge gathering and integration. We define the architecture of a language compiler for processing subject-predicate-object triples using the OpenNLP parser. We implement a depth-first search traversal on the POS tagged syntactic tree appending predicate and object information. A parser enables higher precision and higher recall extractions of syntactic relationships across conjunction boundaries. We are able to extract 2-2.5 times the correct extractions of ReVerb. The extractions are used in a variety of semantic web applications and question answering. We verify extraction of 50,000 triples on the ClueWeb dataset.
https://arxiv.org/abs/1811.05768
The abundance of text data being produced in the modern age makes it increasingly important to intuitively group, categorize, or classify text data by theme for efficient retrieval and search. Yet, the high dimensionality and imprecision of text data, or more generally language as a whole, prove to be challenging when attempting to perform unsupervised document clustering. In this thesis, we present two novel methods for improving unsupervised document clustering/classification by theme. The first is to improve document representations. We look to exploit “term neighborhoods” and “blur” semantic weight across neighboring terms. These neighborhoods are located in the semantic space afforded by “word embeddings.” The second method is for cluster revision, based on what we deem as “stochastic barcoding”, or “S- Barcode” patterns. Text data is inherently high dimensional, yet clustering typically takes place in a low dimensional representation space. Our method utilizes lower dimension clustering results as initial cluster configurations, and iteratively revises the configuration in the high dimensional space. We show with experimental results how both of the two methods improve the quality of document clustering. While this thesis elaborates on the two new conceptual contributions, a joint thesis by David Yan details the feature transformation and software architecture we developed for unsupervised document classification.
https://arxiv.org/abs/1811.02456
We propose a novel and flexible anchor mechanism named MetaAnchor for object detection frameworks. Unlike many previous detectors model anchors via a predefined manner, in MetaAnchor anchor functions could be dynamically generated from the arbitrary customized prior boxes. Taking advantage of weight prediction, MetaAnchor is able to work with most of the anchor-based object detection systems such as RetinaNet. Compared with the predefined anchor scheme, we empirically find that MetaAnchor is more robust to anchor settings and bounding box distributions; in addition, it also shows the potential on transfer tasks. Our experiment on COCO detection task shows that MetaAnchor consistently outperforms the counterparts in various scenarios.
https://arxiv.org/abs/1807.00980
Community detection is one of the most important and interesting issues in social network analysis. In recent years, simultaneous considering of nodes’ attributes and topological structures of social networks in the process of community detection has attracted the attentions of many scholars, and this consideration has been recently used in some community detection methods to increase their efficiencies and to enhance their performances in finding meaningful and relevant communities. But the problem is that most of these methods tend to find non-overlapping communities, while many real-world networks include communities that often overlap to some extent. In order to solve this problem, an evolutionary algorithm called MOBBO-OCD, which is based on multi-objective biogeography-based optimization (BBO), is proposed in this paper to automatically find overlapping communities in a social network with node attributes with synchronously considering the density of connections and the similarity of nodes’ attributes in the network. In MOBBO-OCD, an extended locus-based adjacency representation called OLAR is introduced to encode and decode overlapping communities. Based on OLAR, a rank-based migration operator along with a novel two-phase mutation strategy and a new double-point crossover are used in the evolution process of MOBBO-OCD to effectively lead the population into the evolution path. In order to assess the performance of MOBBO-OCD, a new metric called alpha_SAEM is proposed in this paper, which is able to evaluate the goodness of both overlapping and non-overlapping partitions with considering the two aspects of node attributes and linkage structure. Quantitative evaluations reveal that MOBBO-OCD achieves favorable results which are quite superior to the results of 15 relevant community detection algorithms in the literature.
https://arxiv.org/abs/1811.02309
We frame unsupervised machine translation (MT) in the context of multi-task learning (MTL), combining insights from both directions. We leverage off-the-shelf neural MT architectures to train unsupervised MT models with no parallel data and show that such models can achieve reasonably good performance, competitive with models purpose-built for unsupervised MT. Finally, we propose improvements that allow us to apply our models to English-Turkish, a truly low-resource language pair.
https://arxiv.org/abs/1811.02278
Since its introduction, unsupervised representation learning has attracted a lot of attention from the research community, as it is demonstrated to be highly effective and easy-to-apply in tasks such as dimension reduction, clustering, visualization, information retrieval, and semi-supervised learning. In this work, we propose a novel unsupervised representation learning framework called neighbor-encoder, in which domain knowledge can be easily incorporated into the learning process without modifying the general encoder-decoder architecture of the classic this http URL contrast to autoencoder, which reconstructs the input data itself, neighbor-encoder reconstructs the input data’s neighbors. As the proposed representation learning problem is essentially a neighbor reconstruction problem, domain knowledge can be easily incorporated in the form of an appropriate definition of similarity between objects. Based on that observation, our framework can leverage any off-the-shelf similarity search algorithms or side information to find the neighbor of an input object. Applications of other algorithms (e.g., association rule mining) in our framework are also possible, given that the appropriate definition of neighbor can vary in different contexts. We have demonstrated the effectiveness of our framework in many diverse domains, including images, text, and time series, and for various data mining tasks including classification, clustering, and visualization. Experimental results show that neighbor-encoder not only outperforms autoencoder in most of the scenarios we consider, but also achieves the state-of-the-art performance on text document clustering.
https://arxiv.org/abs/1811.01557
Graph based entropy, an index of the diversity of events in their distribution to parts of a co-occurrence graph, is proposed for detecting signs of structural changes in the data that are informative in explaining latent dynamics of consumers behavior. For obtaining graph-based entropy, connected subgraphs are first obtained from the graph of co-occurrences of items in the data. Then, the distribution of items occurring in events in the data to these sub-graphs is reflected on the value of graph-based entropy. For the data on the position of sale, a change in this value is regarded as a sign of the appearance, the separation, the disappearance, or the uniting of consumers interests. These phenomena are regarded as the signs of dynamic changes in consumers behavior that may be the effects of external events and information. Experiments show that graph-based entropy outperforms baseline methods that can be used for change detection, in explaining substantial changes and their signs in consumers preference of items in supermarket stores.
http://arxiv.org/abs/1811.12165
The rapid evolution of scientific research has been creating a huge volume of publications every year. Among the many quantification measures of scientific impact, citation count stands out for its frequent use in the research community. Although peer review process is the mainly reliable way of predicting a paper’s future impact, the ability to foresee lasting impact on the basis of citation records is increasingly important in the scientific impact analysis in the era of big data. This paper focuses on the long-term citation count prediction for individual publications, which has become an emerging and challenging applied research topic. Based on the four key phenomena confirmed independently in previous studies of long-term scientific impact quantification, including the intrinsic quality of publications, the aging effect and the Matthew effect and the recency effect, we unify the formulations of all these observations in this paper. Building on a foundation of the above formulations, we propose a long-term citation count prediction model for individual papers via recurrent neural network with long short-term memory units. Extensive experiments on a real-large citation data set demonstrate that the proposed model consistently outperforms existing methods, and achieves a significant performance improvement.
https://arxiv.org/abs/1811.02129
Recently Neural Architecture Search (NAS) has aroused great interest in both academia and industry, however it remains challenging because of its huge and non-continuous search space. Instead of applying evolutionary algorithm or reinforcement learning as previous works, this paper proposes a Direct Sparse Optimization NAS (DSO-NAS) method. In DSO-NAS, we provide a novel model pruning view to NAS problem. In specific, we start from a completely connected block, and then introduce scaling factors to scale the information flow between operations. Next, we impose sparse regularizations to prune useless connections in the architecture. Lastly, we derive an efficient and theoretically sound optimization method to solve it. Our method enjoys both advantages of differentiability and efficiency, therefore can be directly applied to large datasets like ImageNet. Particularly, On CIFAR-10 dataset, DSO-NAS achieves an average test error 2.84\%, while on the ImageNet dataset DSO-NAS achieves 25.4\% test error under 600M FLOPs with 8 GPUs in 18 hours.
https://arxiv.org/abs/1811.01567
Most modern neural machine translation (NMT) systems rely on presegmented inputs. Segmentation granularity importantly determines the input and output sequence lengths, hence the modeling depth, and source and target vocabularies, which in turn determine model size, computational costs of softmax normalization, and handling of out-of-vocabulary words. However, the current practice is to use static, heuristic-based segmentations that are fixed before NMT training. This begs the question whether the chosen segmentation is optimal for the translation task. To overcome suboptimal segmentation choices, we present an algorithm for dynamic segmentation based on the Adaptative Computation Time algorithm (Graves 2016), that is trainable end-to-end and driven by the NMT objective. In an evaluation on four translation tasks we found that, given the freedom to navigate between different segmentation levels, the model prefers to operate on (almost) character level, providing support for purely character-level NMT models from a novel angle.
https://arxiv.org/abs/1810.01480
Video object segmentation is a fundamental step in many advanced vision applications. Most existing algorithms are based on handcrafted features such as HOG, super-pixel segmentation or texture-based techniques, while recently deep features have been found to be more efficient. Existing algorithms observe performance degradation in the presence of challenges such as illumination variations, shadows, and color camouflage. To handle these challenges we propose a fusion based moving object segmentation algorithm which exploits color as well as depth information using GAN to achieve more accuracy. Our goal is to segment moving objects in the presence of challenging background scenes, in real environments. To address this problem, GAN is trained in an unsupervised manner on color and depth information independently with challenging video sequences. During testing, the trained GAN generates backgrounds similar to that in the test sample. The generated background samples are then compared with the test sample to segment moving objects. The final result is computed by fusion of object boundaries in both modalities, RGB and the depth. The comparison of our proposed algorithm with five state-of-the-art methods on publicly available dataset has shown the strength of our algorithm for moving object segmentation in videos in the presence of challenging real scenarios.
https://arxiv.org/abs/1811.01526
We propose two new techniques for training Generative Adversarial Networks (GANs). Our objectives are to alleviate mode collapse in GAN and improve the quality of the generated samples. First, we propose neighbor embedding, a manifold learning-based regularization to explicitly retain local structures of latent samples in the generated samples. This prevents generator from producing nearly identical data samples from different latent samples, and reduces mode collapse. We propose an inverse t-SNE regularizer to achieve this. Second, we propose a new technique, gradient matching, to align the distributions of the generated samples and the real samples. As it is challenging to work with high-dimensional sample distributions, we propose to align these distributions through the scalar discriminator scores. We constrain the difference between the discriminator scores of the real samples and generated ones. We further constrain the difference between the gradients of these discriminator scores. We derive these constraints from Taylor approximations of the discriminator function. We perform experiments to demonstrate that our proposed techniques are computationally simple and easy to be incorporated in existing systems. When Gradient matching and Neighbour embedding are applied together, our GN-GAN achieves outstanding results on 1D/2D synthetic, CIFAR-10 and STL-10 datasets, e.g. FID score of $30.80$ for the STL-10 dataset. Our code is available at: this https URL
https://arxiv.org/abs/1811.01333
Because of their effectiveness in broad practical applications, LSTM networks have received a wealth of coverage in scientific journals, technical blogs, and implementation guides. However, in most articles, the inference formulas for the LSTM network and its parent, RNN, are stated axiomatically, while the training formulas are omitted altogether. In addition, the technique of “unrolling” an RNN is routinely presented without justification throughout the literature. The goal of this paper is to explain the essential RNN and LSTM fundamentals in a single document. Drawing from concepts in signal processing, we formally derive the canonical RNN formulation from differential equations. We then propose and prove a precise statement, which yields the RNN unrolling technique. We also review the difficulties with training the standard RNN and address them by transforming the RNN into the “Vanilla LSTM” network through a series of logical arguments. We provide all equations pertaining to the LSTM system together with detailed descriptions of its constituent entities. Albeit unconventional, our choice of notation and the method for presenting the LSTM system emphasizes ease of understanding. As part of the analysis, we identify new opportunities to enrich the LSTM system and incorporate these extensions into the Vanilla LSTM network, producing the most general LSTM variant to date. The target reader has already been exposed to RNNs and LSTM networks through numerous available resources and is open to an alternative pedagogical approach. A Machine Learning practitioner seeking guidance for implementing our new augmented LSTM model in software for experimentation and research will find the insights and derivations in this tutorial valuable as well.
https://arxiv.org/abs/1808.03314
Sorting extremely large datasets is a frequently occuring task in practice. These datasets are usually much larger than the computer’s main memory; thus external memory sorting algorithms, first introduced by Aggarwal and Vitter (1988), are often used. The complexity of comparison based external memory sorting has been understood for decades by now, however the situation remains elusive if we assume the keys to be sorted are integers. In internal memory, one can sort a set of $n$ integer keys of $\Theta(\lg n)$ bits each in $O(n)$ time using the classic Radix Sort algorithm, however in external memory, there are no faster integer sorting algorithms known than the simple comparison based ones. In this paper, we present a tight conditional lower bound on the complexity of external memory sorting of integers. Our lower bound is based on a famous conjecture in network coding by Li and Li, who conjectured that network coding cannot help anything beyond the standard multicommodity flow rate in undirected graphs. The only previous work connecting the Li and Li conjecture to lower bounds for algorithms is due to Adler et al. Adler et al. indeed obtain relatively simple lower bounds for oblivious algorithms (the memory access pattern is fixed and independent of the input data). Unfortunately obliviousness is a strong limitations, especially for integer sorting: we show that the Li and Li conjecture implies an $\Omega(n \log n)$ lower bound for internal memory oblivious sorting when the keys are $\Theta(\lg n)$ bits. This is in sharp contrast to the classic (non-oblivious) Radix Sort algorithm. Indeed going beyond obliviousness is highly non-trivial; we need to introduce several new methods and involved techniques, which are of their own interest, to obtain our tight lower bound for external memory integer sorting.
https://arxiv.org/abs/1811.01313
A longstanding problem in machine learning is to find unsupervised methods that can learn the statistical structure of high dimensional signals. In recent years, GANs have gained much attention as a possible solution to the problem, and in particular have shown the ability to generate remarkably realistic high resolution sampled images. At the same time, many authors have pointed out that GANs may fail to model the full distribution (“mode collapse”) and that using the learned models for anything other than generating samples may be very difficult. In this paper, we examine the utility of GANs in learning statistical models of images by comparing them to perhaps the simplest statistical model, the Gaussian Mixture Model. First, we present a simple method to evaluate generative models based on relative proportions of samples that fall into predetermined bins. Unlike previous automatic methods for evaluating models, our method does not rely on an additional neural network nor does it require approximating intractable computations. Second, we compare the performance of GANs to GMMs trained on the same datasets. While GMMs have previously been shown to be successful in modeling small patches of images, we show how to train them on full sized images despite the high dimensionality. Our results show that GMMs can generate realistic samples (although less sharp than those of GANs) but also capture the full distribution, which GANs fail to do. Furthermore, GMMs allow efficient inference and explicit representation of the underlying statistical structure. Finally, we discuss how GMMs can be used to generate sharp images.
https://arxiv.org/abs/1805.12462
Recently, considerable effort has been devoted to deep domain adaptation in computer vision and machine learning communities. However, most of existing work only concentrates on learning shared feature representation by minimizing the distribution discrepancy across different domains. Due to the fact that all the domain alignment approaches can only reduce, but not remove the domain shift. Target domain samples distributed near the edge of the clusters, or far from their corresponding class centers are easily to be misclassified by the hyperplane learned from the source domain. To alleviate this issue, we propose to joint domain alignment and discriminative feature learning, which could benefit both domain alignment and final classification. Specifically, an instance-based discriminative feature learning method and a center-based discriminative feature learning method are proposed, both of which guarantee the domain invariant features with better intra-class compactness and inter-class separability. Extensive experiments show that learning the discriminative features in the shared feature space can significantly boost the performance of deep domain adaptation methods.
http://arxiv.org/abs/1808.09347
Given an untrimmed video and a sentence description, temporal sentence localization aims to automatically determine the start and end points of the described sentence within the video. The problem is challenging as it needs the understanding of both video and sentence. Existing research predominantly employs a costly “scan and localize” framework, neglecting the global video context and the specific details within sentences which play as critical issues for this problem. In this paper, we propose a novel Attention Based Location Regression (ABLR) approach to solve the temporal sentence localization from a global perspective. Specifically, to preserve the context information, ABLR first encodes both video and sentence via Bidirectional LSTM networks. Then, a multi-modal co-attention mechanism is introduced to generate not only video attention which reflects the global video structure, but also sentence attention which highlights the crucial details for temporal localization. Finally, a novel attention based location regression network is designed to predict the temporal coordinates of sentence query from the previous attention. ABLR is jointly trained in an end-to-end manner. Comprehensive experiments on ActivityNet Captions and TACoS datasets demonstrate both the effectiveness and the efficiency of the proposed ABLR approach.
https://arxiv.org/abs/1804.07014
Neural machine translation (NMT) models learn representations containing substantial linguistic information. However, it is not clear if such information is fully distributed or if some of it can be attributed to individual neurons. We develop unsupervised methods for discovering important neurons in NMT models. Our methods rely on the intuition that different models learn similar properties, and do not require any costly external supervision. We show experimentally that translation quality depends on the discovered neurons, and find that many of them capture common linguistic phenomena. Finally, we show how to control NMT translations in predictable ways, by modifying activations of individual neurons.
https://arxiv.org/abs/1811.01157
Sequential learning of tasks using gradient descent leads to an unremitting decline in the accuracy of tasks for which training data is no longer available, termed catastrophic forgetting. Generative models have been explored as a means to approximate the distribution of old tasks and bypass storage of real data. Here we propose a cumulative closed-loop generator and embedded classifier using an AC-GAN architecture provided with external regularization by a small buffer. We evaluate incremental learning using a notoriously hard paradigm, single headed learning, in which each task is a disjoint subset of classes in the overall dataset, and performance is evaluated on all previous classes. First, we show that the variability contained in a small percentage of a dataset (memory buffer) accounts for a significant portion of the reported accuracy, both in multi-task and continual learning settings. Second, we show that using a generator to continuously output new images while training provides an up-sampling of the buffer, which prevents catastrophic forgetting and yields superior performance when compared to a fixed buffer. We achieve an average accuracy for all classes of 92.26% in MNIST and 76.15% in FASHION-MNIST after 5 tasks using GAN sampling with a buffer of only 0.17% of the entire dataset size. We compare to a network with regularization (EWC) which shows a deteriorated average performance of 29.19% (MNIST) and 26.5% (FASHION). The baseline of no regularization (plain gradient descent) performs at 99.84% (MNIST) and 99.79% (FASHION) for the last task, but below 3% for all previous tasks. Our method has very low long-term memory cost, the buffer, as well as negligible intermediate memory storage.
https://arxiv.org/abs/1811.01146
We propose a method to transfer knowledge across neural machine translation (NMT) models by means of a shared dynamic vocabulary. Our approach allows to extend an initial model for a given language pair to cover new languages by adapting its vocabulary as long as new data become available (i.e., introducing new vocabulary items if they are not included in the initial model). The parameter transfer mechanism is evaluated in two scenarios: i) to adapt a trained single language NMT system to work with a new language pair and ii) to continuously add new language pairs to grow to a multilingual NMT system. In both the scenarios our goal is to improve the translation performance, while minimizing the training convergence time. Preliminary experiments spanning five languages with different training data sizes (i.e., 5k and 50k parallel sentences) show a significant performance gain ranging from +3.85 up to +13.63 BLEU in different language directions. Moreover, when compared with training an NMT model from scratch, our transfer-learning approach allows us to reach higher performance after training up to 4% of the total training steps.
https://arxiv.org/abs/1811.01137
In this work a systematic approach to the search for all-$sp^2$ bonded carbon allotropes with low density is presented. In particular, we obtain a number of novel energetically stable crystal structures, whose arrangement is closely related to the topology of graphene, by modifying the packing of congruent discs under the condition of local stability. Our procedure starts from an initial parent topology and proceeds to generate daughter architectures derived by lowering the packing factors. Furthermore, we assess both the electronic properties, such as the band structure and the density of states, and the mechanical properties, such as the elastic constants and the stress–strain characteristics, of parent’s and daughter’s geometries from first-principle simulations. We find, using geometrical packing arguments, that some arrangements lead to a density as low as half that of graphene, obtaining some of the least dense structures of all-$sp^2$ bonded carbon allotropes that could ever be synthesized. Nevertheless, a threshold value of the density exists below which the mechanical rigidity of graphene is irreparably lost, while keeping other mechanical characteristics, such as the specific toughness and strength, almost unchanged with lower weight.
https://arxiv.org/abs/1811.01112
Despite recent work in Reading Comprehension (RC), progress has been mostly limited to English due to the lack of large-scale datasets in other languages. In this work, we introduce the first RC system for languages without RC training data. Given a target language without RC training data and a pivot language with RC training data (e.g. English), our method leverages existing RC resources in the pivot language by combining a competitive RC model in the pivot language with an attentive Neural Machine Translation (NMT) model. We first translate the data from the target to the pivot language, and then obtain an answer using the RC model in the pivot language. Finally, we recover the corresponding answer in the original language using soft-alignment attention scores from the NMT model. We create evaluation sets of RC data in two non-English languages, namely Japanese and French, to evaluate our method. Experimental results on these datasets show that our method significantly outperforms a back-translation baseline of a state-of-the-art product-level machine translation system.
https://arxiv.org/abs/1809.03275
We present Open Images V4, a dataset of 9.2M images with unified annotations for image classification, object detection and visual relationship detection. The images have a Creative Commons Attribution license that allows to share and adapt the material, and they have been collected from Flickr without a predefined list of class names or tags, leading to natural class statistics and avoiding an initial design bias. Open Images V4 offers large scale across several dimensions: 30.1M image-level labels for 19.8k concepts, 15.4M bounding boxes for 600 object classes, and 375k visual relationship annotations involving 57 classes. For object detection in particular, we provide 15x more bounding boxes than the next largest datasets (15.4M boxes on 1.9M images). The images often show complex scenes with several objects (8 annotated objects per image on average). We annotated visual relationships between them, which support visual relationship detection, an emerging task that requires structured reasoning. We provide in-depth comprehensive statistics about the dataset, we validate the quality of the annotations, and we study how the performance of many modern models evolves with increasing amounts of training data. We hope that the scale, quality, and variety of Open Images V4 will foster further research and innovation even beyond the areas of image classification, object detection, and visual relationship detection.
https://arxiv.org/abs/1811.00982
We propose in this paper a combined model of Long Short Term Memory and Convolutional Neural Networks (LSTM-CNN) that exploits word embeddings and positional embeddings for cross-sentence n-ary relation extraction. The proposed model brings together the properties of both LSTMs and CNNs, to simultaneously exploit long-range sequential information and capture most informative features, essential for cross-sentence n-ary relation extraction. The LSTM-CNN model is evaluated on standard dataset on cross-sentence n-ary relation extraction, where it significantly outperforms baselines such as CNNs, LSTMs and also a combined CNN-LSTM model. The paper also shows that the LSTM-CNN model outperforms the current state-of-the-art methods on cross-sentence n-ary relation extraction.
https://arxiv.org/abs/1811.00845
Although neural machine translation (NMT) has achieved impressive progress recently, it is usually trained on the clean parallel data set and hence cannot work well when the input sentence is the production of the automatic speech recognition (ASR) system due to the enormous errors in the source. To solve this problem, we propose a simple but effective method to improve the robustness of NMT in the case of speech translation. We simulate the noise existing in the realistic output of the ASR system and inject them into the clean parallel data so that NMT can work under similar word distributions during training and testing. Besides, we also incorporate the Chinese Pinyin feature which is easy to get in speech translation to further improve the translation performance. Experiment results show that our method has a more stable performance and outperforms the baseline by an average of 3.12 BLEU on multiple noisy test sets, even while achieves a generalization improvement on the WMT’17 Chinese-English test set.
https://arxiv.org/abs/1811.00728
Acquiring a large vocabulary is an important aspect of human intelligence. Onecommon approach for human to populating vocabulary is to learn words duringreading or listening, and then use them in writing or speaking. This ability totransfer from input to output is natural for human, but it is difficult for machines.Human spontaneously performs this knowledge transfer in complicated multimodaltasks, such as Visual Question Answering (VQA). In order to approach human-levelArtificial Intelligence, we hope to equip machines with such ability. Therefore, toaccelerate this research, we propose a newzero-shot transfer VQA(ZST-VQA)dataset by reorganizing the existing VQA v1.0 dataset in the way that duringtraining, some words appear only in one module (i.e. questions) but not in theother (i.e. answers). In this setting, an intelligent model should understand andlearn the concepts from one module (i.e. questions), and at test time, transfer themto the other (i.e. predict the concepts as answers). We conduct evaluation on thisnew dataset using three existing state-of-the-art VQA neural models. Experimentalresults show a significant drop in performance on this dataset, indicating existingmethods do not address the zero-shot transfer problem. Besides, our analysis findsthat this may be caused by the implicit bias learned during training.
https://arxiv.org/abs/1811.00692
In this paper, we propose a multilingual encoder-decoder architecture capable of obtaining multilingual sentence representations by means of incorporating an intermediate {\em attention bridge} that is shared across all languages. That is, we train the model with language-specific encoders and decoders that are connected via self-attention with a shared layer that we call attention bridge. This layer exploits the semantics from each language for performing translation and develops into a language-independent meaning representation that can efficiently be used for transfer learning. We present a new framework for the efficient development of multilingual NMT using this model and scheduled training. We have tested the approach in a systematic way with a multi-parallel data set. We show that the model achieves substantial improvements over strong bilingual models and that it also works well for zero-shot translation, which demonstrates its ability of abstraction and transfer learning.
https://arxiv.org/abs/1811.00498
We introduce a new dataset for joint reasoning about language and vision. The data contains 107,296 examples of English sentences paired with web photographs. The task is to determine whether a natural language caption is true about a photograph. We present an approach for finding visually complex images and crowdsourcing linguistically diverse captions. Qualitative analysis shows the data requires complex reasoning about quantities, comparisons, and relationships between objects. Evaluation of state-of-the-art visual reasoning methods shows the data is a challenge for current methods.
https://arxiv.org/abs/1811.00491
Visual Question answering is a challenging problem requiring a combination of concepts from Computer Vision and Natural Language Processing. Most existing approaches use a two streams strategy, computing image and question features that are consequently merged using a variety of techniques. Nonetheless, very few rely on higher level image representations, which can capture semantic and spatial relationships. In this paper, we propose a novel graph-based approach for Visual Question Answering. Our method combines a graph learner module, which learns a question specific graph representation of the input image, with the recent concept of graph convolutions, aiming to learn image representations that capture question specific interactions. We test our approach on the VQA v2 dataset using a simple baseline architecture enhanced by the proposed graph learner module. We obtain promising results with 66.18% accuracy and demonstrate the interpretability of the proposed method. Code can be found at github.com/aimbrain/vqa-project.
https://arxiv.org/abs/1806.07243
Image quality measurement is a critical problem for image super-resolution (SR) algorithms. Usually, they are evaluated by some well-known objective metrics, e.g., PSNR and SSIM, but these indices cannot provide suitable results in accordance with the perception of human being. Recently, a more reasonable perception measurement has been proposed in [1], which is also adopted by the PIRM-SR 2018 challenge. In this paper, motivated by [1], we aim to generate a high-quality SR result which balances between the two indices, i.e., the perception index and root-mean-square error (RMSE). To do so, we design a new deep SR framework, dubbed Bi-GANs-ST, by integrating two complementary generative adversarial networks (GAN) branches. One is memory residual SRGAN (MR-SRGAN), which emphasizes on improving the objective performance, such as reducing the RMSE. The other is weight perception SRGAN (WP-SRGAN), which obtains the result that favors better subjective perception via a two-stage adversarial training mechanism. Then, to produce final result with excellent perception scores and RMSE, we use soft-thresholding method to merge the results generated by the two GANs. Our method performs well on the perceptual image super-resolution task of the PIRM 2018 challenge. Experimental results on five benchmarks show that our proposal achieves highly competent performance compared with other state-of-the-art methods.
https://arxiv.org/abs/1811.00367
Automatic multi-class object detection in remote sensing images in unconstrained scenarios is of high interest for several applications including traffic monitoring and disaster management. The huge variation in object scale, orientation, category, and complex backgrounds, as well as the different camera sensors pose great challenges for current algorithms. In this work, we propose a new method consisting of a novel joint image cascade and feature pyramid network with multi-size convolution kernels to extract multi-scale strong and weak semantic features. These features are fed into rotation-based region proposal and region of interest networks to produce object detections. Finally, rotational non-maximum suppression is applied to remove redundant detections. During training, we minimize joint horizontal and oriented bounding box loss functions, as well as a novel loss that enforces oriented boxes to be rectangular. Our method achieves 68.16% mAP on horizontal and 72.45% mAP on oriented bounding box detection tasks on the challenging DOTA dataset, outperforming all published methods by a large margin (+6% and +12% absolute improvement, respectively). Furthermore, it generalizes to two other datasets, NWPU VHR-10 and UCAS-AOD, and achieves competitive results with the baselines even when trained on DOTA. Our method can be deployed in multi-class object detection applications, regardless of the image and object scales and orientations, making it a great choice for unconstrained aerial and satellite imagery.
https://arxiv.org/abs/1807.02700
In this study, we first investigate a novel capsule network with dynamic routing for linear time Neural Machine Translation (NMT), referred as \textsc{CapsNMT}. \textsc{CapsNMT} uses an aggregation mechanism to map the source sentence into a matrix with pre-determined size, and then applys a deep LSTM network to decode the target sequence from the source representation. Unlike the previous work \cite{sutskever2014sequence} to store the source sentence with a passive and bottom-up way, the dynamic routing policy encodes the source sentence with an iterative process to decide the credit attribution between nodes from lower and higher layers. \textsc{CapsNMT} has two core properties: it runs in time that is linear in the length of the sequences and provides a more flexible way to select, represent and aggregates the part-whole information of the source sentence. On WMT14 English-German task and a larger WMT14 English-French task, \textsc{CapsNMT} achieves comparable results with the state-of-the-art NMT systems. To the best of our knowledge, this is the first work that capsule networks have been empirically investigated for sequence to sequence problems.
https://arxiv.org/abs/1811.00287
In this paper, we propose the multi-domain dictionary learning (MDDL) to make dictionary learning-based classification more robust to data representing in different domains. We use adversarial neural networks to generate data in different styles, and collect all the generated data into a miscellaneous dictionary. To tackle the dictionary learning with many samples, we compute the weighting matrix that compress the miscellaneous dictionary from multi-sample per class to single sample per class. We show that the time complexity solving the proposed MDDL with weighting matrix is the same as solving the dictionary with single sample per class. Moreover, since the weighting matrix could help the solver rely more on the training data, which possibly lie in the same domain with the testing data, the classification could be more accurate.
https://arxiv.org/abs/1811.00274
Current Neural Machine Translation (NMT) employs a language-specific encoder to represent the source sentence and adopts a language-specific decoder to generate target translation. This language-dependent design leads to large-scale network parameters and makes the duality of the parallel data underutilized. To address the problem, we propose in this paper a language-independent representor to replace the encoder and decoder by using weight sharing. This shared representor can not only reduce large portion of network parameters, but also facilitate us to fully explore the language duality by jointly training source-to-target, target-to-source, left-to-right and right-to-left translations within a multi-task learning framework. Experiments show that our proposed framework can obtain significant improvements over conventional NMT models on resource-rich and low-resource translation tasks with only a quarter of parameters.
https://arxiv.org/abs/1811.00258
This paper proposes a new architecture - Attentive Tensor Product Learning (ATPL) - to represent grammatical structures in deep learning models. ATPL is a new architecture to bridge this gap by exploiting Tensor Product Representations (TPR), a structured neural-symbolic model developed in cognitive science, aiming to integrate deep learning with explicit language structures and rules. The key ideas of ATPL are: 1) unsupervised learning of role-unbinding vectors of words via TPR-based deep neural network; 2) employing attention modules to compute TPR; and 3) integration of TPR with typical deep learning architectures including Long Short-Term Memory (LSTM) and Feedforward Neural Network (FFNN). The novelty of our approach lies in its ability to extract the grammatical structure of a sentence by using role-unbinding vectors, which are obtained in an unsupervised manner. This ATPL approach is applied to 1) image captioning, 2) part of speech (POS) tagging, and 3) constituency parsing of a sentence. Experimental results demonstrate the effectiveness of the proposed approach.
https://arxiv.org/abs/1802.07089
Most counting questions in visual question answering (VQA) datasets are simple and require no more than object detection. Here, we study algorithms for complex counting questions that involve relationships between objects, attribute identification, reasoning, and more. To do this, we created TallyQA, the world’s largest dataset for open-ended counting. We propose a new algorithm for counting that uses relation networks with region proposals. Our method lets relation networks be efficiently used with high-resolution imagery. It yields state-of-the-art results compared to baseline and recent systems on both TallyQA and the HowMany-QA benchmark.
https://arxiv.org/abs/1810.12440
Concept-to-text generation typically employs a pipeline architecture, which often leads to suboptimal texts. Content selection, for example, may greedily select the most important facts, which may require, however, too many words to express, and this may be undesirable when space is limited or expensive. Selecting other facts, possibly only slightly less important, may allow the lexicalization stage to use much fewer words, or to report more facts in the same space. Decisions made during content selection and lexicalization may also lead to more or fewer sentence aggregation opportunities, affecting the length and readability of the resulting texts. Building upon on a publicly available state of the art natural language generator for Semantic Web ontologies, this article presents an Integer Linear Programming model that, unlike pipeline architectures, jointly considers choices available in content selection, lexicalization, and sentence aggregation to avoid greedy local decisions and produce more compact texts, i.e., texts that report more facts per word. Compact texts are desirable, for example, when generating advertisements to be included in Web search results, or when summarizing structured information in limited space. An extended version of the proposed model also considers a limited form of referring expression generation and avoids redundant sentences. An approximation of the two models can be used when longer texts need to be generated. Experiments with three ontologies confirm that the proposed models lead to more compact texts, compared to pipeline systems, with no deterioration or with improvements in the perceived quality of the generated texts.
https://arxiv.org/abs/1811.00051
Gate-based universal quantum computers form a rapidly evolving field of quantum computing hardware technology. In previous work, we presented a quantum algorithm for lattice protein folding on a cubic lattice, tailored for quantum annealers. In this paper, we introduce a novel approach for solving the lattice protein folding problem on universal gate-based quantum computing architectures. Lattice protein models are coarse-grained representations of proteins that have been used extensively over the past thirty years to examine the principles of protein folding and design.These models can be used to explore a vast number of possible protein conformations and to infer structural properties of more complex atomistic protein structures. We formulate the problem as a quantum alternating operator ansatz, a member of the wider class of variational quantum/classical hybrid algorithms. To increase the probability of sampling the ground state, we propose splitting the optimization problem into hard and soft constraints. This enables us to use a previously under-utilised component of the variational algorithm to constrain the search to the subspace of solutions that satisfy the hard constraints.
https://arxiv.org/abs/1810.13411
In NMT, how far can we get without attention and without separate encoding and decoding? To answer that question, we introduce a recurrent neural translation model that does not use attention and does not have a separate encoder and decoder. Our eager translation model is low-latency, writing target tokens as soon as it reads the first source token, and uses constant memory during decoding. It performs on par with the standard attention-based model of Bahdanau et al. (2014), and better on long sentences.
https://arxiv.org/abs/1810.13409