arXiv subjects

Yi Yang

Publications and source records attributed to Yi Yang.

At least 1,063 records · Page 59Linked to original sources

arXiv2017

An Improved Residual LSTM Architecture for Acoustic Modeling

Long Short-Term Memory (LSTM) is the primary recurrent neural networks architecture for acoustic modeling in automatic speech recognition systems. Residual learning is an efficient method to help neural networks converge easier and faster. In this paper, we propose several types of residual LSTM methods for our acoustic modeling. Our experiments indicate that, compared with classic LSTM, our architecture shows more than 8% relative reduction in Phone Error Rate (PER) on TIMIT tasks. At the same time, our residual fast LSTM approach shows 4% relative reduction in PER on the same task. Besides, we find that all this architecture could have good results on THCHS-30, Librispeech and Switchboard corpora.

cs.CL↗

arXiv2017

Solving Lauricella String Scattering Amplitudes through Recurrence Relations

We show that there exist infinite number of recurrence relations valid for all energies among the open bosonic string scattering amplitudes (SSA) of three tachyons and one arbitrary string state, or the Lauricella SSA. Moreover, these infinite number of recurrence relations can be used to solve all the Lauricella SSA and express them in terms of one single four tachyon amplitude. These results extend the solvability of SSA at the high energy, fixed angle scattering limit and those at the Regge scattering limit discovered previously.

hep-th↗

arXiv2017

PatchShuffle Regularization

This paper focuses on regularizing the training of the convolutional neural network (CNN). We propose a new regularization approach named ``PatchShuffle`` that can be adopted in any classification-oriented CNN models. It is easy to implement: in each mini-batch, images or feature maps are randomly chosen to undergo a transformation such that pixels within each local patch are shuffled. Through generating images and feature maps with interior orderless patches, PatchShuffle creates rich local variations, reduces the risk of network overfitting, and can be viewed as a beneficial supplement to various kinds of training regularization techniques, such as weight decay, model ensemble and dropout. Experiments on four representative classification datasets show that PatchShuffle improves the generalization ability of CNN especially when the data is scarce. Moreover, we empirically illustrate that CNN models trained with PatchShuffle are more robust to noise and local changes in an image.

cs.CV↗

arXiv2017

UTS submission to Google YouTube-8M Challenge 2017

In this paper, we present our solution to Google YouTube-8M Video Classification Challenge 2017. We leveraged both video-level and frame-level features in the submission. For video-level classification, we simply used a 200-mixture Mixture of Experts (MoE) layer, which achieves GAP 0.802 on the validation set with a single model. For frame-level classification, we utilized several variants of recurrent neural networks, sequence aggregation with attention mechanism and 1D convolutional models. We achieved GAP 0.8408 on the private testing set with the ensemble model. The source code of our models can be found in \url{https://github.com/ffmpbgrnn/yt8m}.

cs.CV↗

arXiv2017

Pedestrian Alignment Network for Large-scale Person Re-identification

Person re-identification (person re-ID) is mostly viewed as an image retrieval problem. This task aims to search a query person in a large image pool. In practice, person re-ID usually adopts automatic detectors to obtain cropped pedestrian images. However, this process suffers from two types of detector errors: excessive background and part missing. Both errors deteriorate the quality of pedestrian alignment and may compromise pedestrian matching due to the position and scale variances. To address the misalignment problem, we propose that alignment can be learned from an identification procedure. We introduce the pedestrian alignment network (PAN) which allows discriminative embedding learning and pedestrian alignment without extra annotations. Our key observation is that when the convolutional neural network (CNN) learns to discriminate between different identities, the learned feature maps usually exhibit strong activations on the human body rather than the background. The proposed network thus takes advantage of this attention mechanism to adaptively locate and align pedestrians within a bounding box. Visual examples show that pedestrians are better aligned with PAN. Experiments on three large-scale re-ID datasets confirm that PAN improves the discriminative ability of the feature embeddings and yields competitive accuracy with the state-of-the-art methods.

cs.CV↗

arXiv2017

Unsupervised Person Re-identification: Clustering and Fine-tuning

The superiority of deeply learned pedestrian representations has been reported in very recent literature of person re-identification (re-ID). In this paper, we consider the more pragmatic issue of learning a deep feature with no or only a few labels. We propose a progressive unsupervised learning (PUL) method to transfer pretrained deep representations to unseen domains. Our method is easy to implement and can be viewed as an effective baseline for unsupervised re-ID feature learning. Specifically, PUL iterates between 1) pedestrian clustering and 2) fine-tuning of the convolutional neural network (CNN) to improve the original model trained on the irrelevant labeled dataset. Since the clustering results can be very noisy, we add a selection operation between the clustering and fine-tuning. At the beginning when the model is weak, CNN is fine-tuned on a small amount of reliable examples which locate near to cluster centroids in the feature space. As the model becomes stronger in subsequent iterations, more images are being adaptively selected as CNN training samples. Progressively, pedestrian clustering and the CNN model are improved simultaneously until algorithm convergence. This process is naturally formulated as self-paced learning. We then point out promising directions that may lead to further improvement. Extensive experiments on three large-scale re-ID datasets demonstrate that PUL outputs discriminative features that improve the re-ID accuracy.

cs.CV↗

arXiv2017

All-angle negative refraction of highly squeezed plasmon and phonon polaritons in graphene-boron nitride heterostructures

A fundamental building block for nanophotonics is the ability to achieve negative refraction of polaritons, because this could enable the demonstration of many unique nanoscale applications such as deep-subwavelength imaging, superlens, and novel guiding. However, to achieve negative refraction of highly squeezed polaritons, such as plasmon polaritons in graphene and phonon polaritons in boron nitride (BN) with their wavelengths squeezed by a factor over 100, requires the ability to flip the sign of their group velocity at will, which is challenging. Here we reveal that the strong coupling between plasmon and phonon polaritons in graphene-BN heterostructures can be used to flip the sign of the group velocity of the resulting hybrid (plasmon-phonon-polariton) modes. We predict all-angle negative refraction between plasmon and phonon polaritons, and even more surprisingly, between hybrid graphene plasmons, and between hybrid phonon polaritons. Graphene-BN heterostructures thus provide a versatile platform for the design of nano-metasurfaces and nano-imaging elements.

physics.optics↗

arXiv2017

Unsupervised Learning Layers for Video Analysis

This paper presents two unsupervised learning layers (UL layers) for label-free video analysis: one for fully connected layers, and the other for convolutional ones. The proposed UL layers can play two roles: they can be the cost function layer for providing global training signal; meanwhile they can be added to any regular neural network layers for providing local training signals and combined with the training signals backpropagated from upper layers for extracting both slow and fast changing features at layers of different depths. Therefore, the UL layers can be used in either pure unsupervised or semi-supervised settings. Both a closed-form solution and an online learning algorithm for two UL layers are provided. Experiments with unlabeled synthetic and real-world videos demonstrated that the neural networks equipped with UL layers and trained with the proposed online learning algorithm can extract shape and motion information from video sequences of moving objects. The experiments demonstrated the potential applications of UL layers and online learning algorithm to head orientation estimation and moving object localization.

cs.LG↗

arXiv2017

SIFT Meets CNN: A Decade Survey of Instance Retrieval

In the early days, content-based image retrieval (CBIR) was studied with global features. Since 2003, image retrieval based on local descriptors (de facto SIFT) has been extensively studied for over a decade due to the advantage of SIFT in dealing with image transformations. Recently, image representations based on the convolutional neural network (CNN) have attracted increasing interest in the community and demonstrated impressive performance. Given this time of rapid evolution, this article provides a comprehensive survey of instance retrieval over the last decade. Two broad categories, SIFT-based and CNN-based methods, are presented. For the former, according to the codebook size, we organize the literature into using large/medium-sized/small codebooks. For the latter, we discuss three lines of methods, i.e., using pre-trained or fine-tuned CNN models, and hybrid methods. The first two perform a single-pass of an image to the network, while the last category employs a patch-based feature extraction scheme. This survey presents milestones in modern instance retrieval, reviews a broad selection of previous works in different categories, and provides insights on the connection between SIFT and CNN-based methods. After analyzing and comparing retrieval performance of different categories on several datasets, we discuss promising directions towards generic and specialized instance retrieval.

cs.CV↗

arXiv2017

Universal Behaviors of Speed of Sound from Holography

We investigate the speed of sound in $(d+1)$-dimensional field theory by studying its dual $(d+2)$-dimensional gravity theory from gauge/gravity correspondence. Instead of the well known conformal limit $c_s^2 \rightarrow 1/d$ at high temperature, we reveal two more universal quantities in various limits: $c_s^2 \rightarrow (d-1)/16π$ at low temperature and $c_s^2 \rightarrow (d-1)/16πd$ at large chemical potential.

hep-th↗

arXiv2017

More is Less: A More Complicated Network with Less Inference Complexity

In this paper, we present a novel and general network structure towards accelerating the inference process of convolutional neural networks, which is more complicated in network structure yet with less inference complexity. The core idea is to equip each original convolutional layer with another low-cost collaborative layer (LCCL), and the element-wise multiplication of the ReLU outputs of these two parallel layers produces the layer-wise output. The combined layer is potentially more discriminative than the original convolutional layer, and its inference is faster for two reasons: 1) the zero cells of the LCCL feature maps will remain zero after element-wise multiplication, and thus it is safe to skip the calculation of the corresponding high-cost convolution in the original convolutional layer, 2) LCCL is very fast if it is implemented as a 1*1 convolution or only a single filter shared by all channels. Extensive experiments on the CIFAR-10, CIFAR-100 and ILSCRC-2012 benchmarks show that our proposed network structure can accelerate the inference process by 32\% on average with negligible performance drop.

cs.CV↗

arXiv2017

GeneGAN: Learning Object Transfiguration and Attribute Subspace from Unpaired Data

Object Transfiguration replaces an object in an image with another object from a second image. For example it can perform tasks like "putting exactly those eyeglasses from image A on the nose of the person in image B". Usage of exemplar images allows more precise specification of desired modifications and improves the diversity of conditional image generation. However, previous methods that rely on feature space operations, require paired data and/or appearance models for training or disentangling objects from background. In this work, we propose a model that can learn object transfiguration from two unpaired sets of images: one set containing images that "have" that kind of object, and the other set being the opposite, with the mild constraint that the objects be located approximately at the same place. For example, the training data can be one set of reference face images that have eyeglasses, and another set of images that have not, both of which spatially aligned by face landmarks. Despite the weak 0/1 labels, our model can learn an "eyeglasses" subspace that contain multiple representatives of different types of glasses. Consequently, we can perform fine-grained control of generated images, like swapping the glasses in two images by swapping the projected components in the "eyeglasses" subspace, to create novel images of people wearing eyeglasses. Overall, our deterministic generative model learns disentangled attribute subspaces from weakly labeled data by adversarial training. Experiments on CelebA and Multi-PIE datasets validate the effectiveness of the proposed model on real world data, in generating images with specified eyeglasses, smiling, hair styles, and lighting conditions etc. The code is available online.

cs.CV↗

arXiv2017

Performance Assessment of High-dimensional Variable Identification

Since model selection is ubiquitous in data analysis, reproducibility of statistical results demands a serious evaluation of reliability of the employed model selection method, no matter what label it may have in terms of good properties. Instability measures have been proposed for evaluating model selection uncertainty. However, low instability does not necessarily indicate that the selected model is trustworthy, since low instability can also arise when a certain method tends to select an overly parsimonious model. F- and G-measures have become increasingly popular for assessing variable selection performance in theoretical studies and simulation results. However, they are not computable in practice. In this work, we propose an estimation method for F- and G-measures and prove their desirable properties of uniform consistency. This gives the data analyst a valuable tool to compare different variable selection methods based on the data at hand. Extensive simulations are conducted to show the very good finite sample performance of our approach. We further demonstrate the application of our methods using several micro-array gene expression data sets, with intriguing findings.

stat.ME↗

arXiv2017

Optical Sky Brightness and Transparency During the Winter Season at Dome A Antarctica From the Gattini-Allsky Camera

The summit of the Antarctic plateau, Dome A, is proving to be an excellent site for optical, NIR, and THz astronomical observations. GATTINI was a wide-field camera installed on the PLATO instrument module as part of the Chinese-led traverse to Dome A in January, 2009. We present here the measurements of sky brightness with the Gattini ultra-large field of view (90 deg x 90 deg) in the photometric B-, V-, and R-bands, cloud cover statistics measured during the 2009 winter season, and an estimate of the sky transparency. A cumulative probability distribution indicates that the darkest 10% of the nights at Dome A have sky brightness of S_B = 22.98, S_V = 21.86, and S_R = 21.68 mag arcsec^{-2}. These values were obtained around the year 2009 with minimum aurora, and they are comparable to the faintest sky brightness at Mauna Kea and the best sites of northern Chile. Since every filter includes strong auroral lines that effectively contaminate the sky brightness measurements, for instruments working around the auroral lines, either with custom filters or with high spectral resolution instruments, these values could be easily obtained on a more routine basis. In addition, we present example light curves for bright targets to emphasize the unprecedented observational window function available from this ground-based site. These light curves will be published in a future paper.

astro-ph.IM↗

arXiv2017

Long-Term Identity-Aware Multi-Person Tracking for Surveillance Video Summarization

Multi-person tracking plays a critical role in the analysis of surveillance video. However, most existing work focus on shorter-term (e.g. minute-long or hour-long) video sequences. Therefore, we propose a multi-person tracking algorithm for very long-term (e.g. month-long) multi-camera surveillance scenarios. Long-term tracking is challenging because 1) the apparel/appearance of the same person will vary greatly over multiple days and 2) a person will leave and re-enter the scene numerous times. To tackle these challenges, we leverage face recognition information, which is robust to apparel change, to automatically reinitialize our tracker over multiple days of recordings. Unfortunately, recognized faces are unavailable oftentimes. Therefore, our tracker propagates identity information to frames without recognized faces by uncovering the appearance and spatial manifold formed by person detections. We tested our algorithm on a 23-day 15-camera data set (4,935 hours total), and we were able to localize a person 53.2% of the time with 69.8% precision. We further performed video summarization experiments based on our tracking output. Results on 116.25 hours of video showed that we were able to generate a reasonable visual diary (i.e. a summary of what a person did) for different people, thus potentially opening the door to automatic summarization of the vast amount of surveillance video generated every day.

cs.CV↗

arXiv2017

The shadow knows: using shadows to investigate the structure of the pretransitional disk of HD 100453

We present GPI polarized intensity imagery of HD 100453 in Y-, J-, and K1 bands which reveals an inner gap ($9 - 18$ au), an outer disk ($18-39$ au) with two prominent spiral arms, and two azimuthally-localized dark features also present in SPHERE total intensity images (Wagner 2015). SED fitting further suggests the radial gap extends to $1$ au. The narrow, wedge-like shape of the dark features appears similar to predictions of shadows cast by a inner disk which is misaligned with respect to the outer disk. Using the Monte Carlo radiative transfer code HOCHUNCK3D (Whitney 2013), we construct a model of the disk which allows us to determine its physical properties in more detail. From the angular separation of the features we measure the difference in inclination between the disks 45$^{\circ}$, and their major axes, PA = 140$^{\circ}$ east of north for the outer disk and 100$^{\circ}$for the inner disk. We find an outer disk inclination of $25 \pm 10^{\circ}$ from face-on in broad agreement with the Wagner 2015 measurement of 34$^{\circ}$. SPHERE data in J- and H-bands indicate a reddish disk which points to HD 100453 evolving into a young debris disk.

astro-ph.EP↗

arXiv2017

Person Re-identification in the Wild

We present a novel large-scale dataset and comprehensive baselines for end-to-end pedestrian detection and person recognition in raw video frames. Our baselines address three issues: the performance of various combinations of detectors and recognizers, mechanisms for pedestrian detection to help improve overall re-identification accuracy and assessing the effectiveness of different detectors for re-identification. We make three distinct contributions. First, a new dataset, PRW, is introduced to evaluate Person Re-identification in the Wild, using videos acquired through six synchronized cameras. It contains 932 identities and 11,816 frames in which pedestrians are annotated with their bounding box positions and identities. Extensive benchmarking results are presented on this dataset. Second, we show that pedestrian detection aids re-identification through two simple yet effective improvements: a discriminatively trained ID-discriminative Embedding (IDE) in the person subspace using convolutional neural network (CNN) features and a Confidence Weighted Similarity (CWS) metric that incorporates detection scores into similarity measurement. Third, we derive insights in evaluating detector performance for the particular scenario of accurate person re-identification.

cs.CV↗

arXiv2017

Accelerating Deep Neural Network Training with Inconsistent Stochastic Gradient Descent

SGD is the widely adopted method to train CNN. Conceptually it approximates the population with a randomly sampled batch; then it evenly trains batches by conducting a gradient update on every batch in an epoch. In this paper, we demonstrate Sampling Bias, Intrinsic Image Difference and Fixed Cycle Pseudo Random Sampling differentiate batches in training, which then affect learning speeds on them. Because of this, the unbiased treatment of batches involved in SGD creates improper load balancing. To address this issue, we present Inconsistent Stochastic Gradient Descent (ISGD) to dynamically vary training effort according to learning statuses on batches. Specifically ISGD leverages techniques in Statistical Process Control to identify a undertrained batch. Once a batch is undertrained, ISGD solves a new subproblem, a chasing logic plus a conservative constraint, to accelerate the training on the batch while avoid drastic parameter changes. Extensive experiments on a variety of datasets demonstrate ISGD converges faster than SGD. In training AlexNet, ISGD is 21.05\% faster than SGD to reach 56\% top1 accuracy under the exactly same experiment setup. We also extend ISGD to work on multiGPU or heterogeneous distributed system based on data parallelism, enabling the batch size to be the key to scalability. Then we present the study of ISGD batch size to the learning rate, parallelism, synchronization cost, system saturation and scalability. We conclude the optimal ISGD batch size is machine dependent. Various experiments on a multiGPU system validate our claim. In particular, ISGD trains AlexNet to 56.3% top1 and 80.1% top5 accuracy in 11.5 hours with 4 NVIDIA TITAN X at the batch size of 1536.

cs.LG↗