We propose Scene Graph Auto-Encoder (SGAE) that incorporates the language inductive bias into the encoder-decoder image captioning framework for more human-like captions. Intuitively, we humans use the inductive bias to compose collocations and contextual inference in discourse. For example, when we see the relation person on bike', it is natural to replace
on’ with ride' and infer
person riding bike on a road’ even the `road’ is not evident. Therefore, exploiting such bias as a language prior is expected to help the conventional encoder-decoder models less likely overfit to the dataset bias and focus on reasoning. Specifically, we use the scene graph — a directed graph ($\mathcal{G}$) where an object node is connected by adjective nodes and relationship nodes — to represent the complex structural layout of both image ($\mathcal{I}$) and sentence ($\mathcal{S}$). In the textual domain, we use SGAE to learn a dictionary ($\mathcal{D}$) that helps to reconstruct sentences in the $\mathcal{S}\rightarrow \mathcal{G} \rightarrow \mathcal{D} \rightarrow \mathcal{S}$ pipeline, where $\mathcal{D}$ encodes the desired language prior; in the vision-language domain, we use the shared $\mathcal{D}$ to guide the encoder-decoder in the $\mathcal{I}\rightarrow \mathcal{G}\rightarrow \mathcal{D} \rightarrow \mathcal{S}$ pipeline. Thanks to the scene graph representation and shared dictionary, the inductive bias is transferred across domains in principle. We validate the effectiveness of SGAE on the challenging MS-COCO image captioning benchmark, e.g., our SGAE-based single-model achieves a new state-of-the-art $127.8$ CIDEr-D on the Karpathy split, and a competitive $125.5$ CIDEr-D (c40) on the official server even compared to other ensemble models.
https://arxiv.org/abs/1812.02378
Fully automating machine learning pipelines is one of the key challenges of current artificial intelligence research, since practical machine learning often requires costly and time-consuming human-powered processes such as model design, algorithm development, and hyperparameter tuning. In this paper, we verify that automated architecture search synergizes with the effect of gradient-based meta learning. We adopt the progressive neural architecture search \cite{liu:pnas_google:DBLP:journals/corr/abs-1712-00559} to find optimal architectures for meta-learners. The gradient based meta-learner whose architecture was automatically found achieved state-of-the-art results on the 5-shot 5-way Mini-ImageNet classification problem with $74.65\%$ accuracy, which is $11.54\%$ improvement over the result obtained by the first gradient-based meta-learner called MAML \cite{finn:maml:DBLP:conf/icml/FinnAL17}. To our best knowledge, this work is the first successful neural architecture search implementation in the context of meta learning.
https://arxiv.org/abs/1806.06927
Dense event captioning aims to detect and describe all events of interest contained in a video. Despite the advanced development in this area, existing methods tackle this task by making use of dense temporal annotations, which is dramatically source-consuming. This paper formulates a new problem: weakly supervised dense event captioning, which does not require temporal segment annotations for model training. Our solution is based on the one-to-one correspondence assumption, each caption describes one temporal segment, and each temporal segment has one caption, which holds in current benchmark datasets and most real-world cases. We decompose the problem into a pair of dual problems: event captioning and sentence localization and present a cycle system to train our model. Extensive experimental results are provided to demonstrate the ability of our model on both dense event captioning and sentence localization in videos.
https://arxiv.org/abs/1812.03849
The encoder-decoder is the typical framework for Neural Machine Translation (NMT), and different structures have been developed for improving the translation performance. Transformer is one of the most promising structures, which can leverage the self-attention mechanism to capture the semantic dependency from global view. However, it cannot distinguish the relative position of different tokens very well, such as the tokens located at the left or right of the current token, and cannot focus on the local information around the current token either. To alleviate these problems, we propose a novel attention mechanism named Hybrid Self-Attention Network (HySAN) which accommodates some specific-designed masks for self-attention network to extract various semantic, such as the global/local information, the left/right part context. Finally, a squeeze gate is introduced to combine different kinds of SANs for fusion. Experimental results on three machine translation tasks show that our proposed framework outperforms the Transformer baseline significantly and achieves superior results over state-of-the-art NMT systems.
https://arxiv.org/abs/1811.00253
Accurately estimating the remaining useful life (RUL) of industrial machinery is beneficial in many real-world applications. Estimation techniques have mainly utilized linear models or neural network based approaches with a focus on short term time dependencies. This paper, introduces a system model that incorporates temporal convolutions with both long term and short term time dependencies. The proposed network learns salient features and complex temporal variations in sensor values, and predicts the RUL. A data augmentation method is used for increased accuracy. The proposed method is compared with several state-of-the-art algorithms on publicly available datasets. It demonstrates promising results, with superior results for datasets obtained from complex environments.
https://arxiv.org/abs/1810.05644
Neural networks are used in a wide range of applications, such as speech recognition and image processing. There is a strong motivation to improve the performance of these applications due to their industrial and commercial significance. Recently, binary neural networks have shown impressive efficiency and accuracy on image recognition data sets. The nature of the operations performed in these algorithms lend themselves well to specialized hardware and processing-in-memory (PIM) approaches. In this paper, we introduce a spintronic, reconfigurable in-memory accelerator for binary neural networks, NV-Net. NV-Net is capable of being used as a standard STT-MRAM array and a computational substrate simultaneously and allows for massively parallel and energy efficient computation. We evaluate NV-Net using multiple image classifiers and a genomics kernel for similarity matching. Our simulation results show that NV-Net is more energy efficient than alternative CPU, GPU, and FPGA based implementations and is also capable of a higher throughput.
https://arxiv.org/abs/1812.03989
E-commerce sponsored search contributes an important part of revenue for the e-commerce company. In consideration of effectiveness and efficiency, a large-scale sponsored search system commonly adopts a multi-stage architecture. We name these stages as ad retrieval, ad pre-ranking and ad ranking. Ad retrieval and ad pre-ranking are collectively referred to as ad matching in this paper. We propose an end-to-end neural matching framework (EENMF) to model two tasks—vector-based ad retrieval and neural networks based ad pre-ranking. Under the deep matching framework, vector-based ad retrieval harnesses user recent behavior sequence to retrieve relevant ad candidates without the constraint of keyword bidding. Simultaneously, the deep model is employed to perform the global pre-ranking of ad candidates from multiple retrieval paths effectively and efficiently. Besides, the proposed model tries to optimize the pointwise cross-entropy loss which is consistent with the objective of predict models in the ranking stage. We conduct extensive evaluation to validate the performance of the proposed framework. In the real traffic of a large-scale e-commerce sponsored search, the proposed approach significantly outperforms the baseline.
https://arxiv.org/abs/1812.01190
Generative Adversarial Networks (GANs) have recently achieved impressive results for many real-world applications, and many GAN variants have emerged with improvements in sample quality and training stability. However, they have not been well visualized or understood. How does a GAN represent our visual world internally? What causes the artifacts in GAN results? How do architectural choices affect GAN learning? Answering such questions could enable us to develop new insights and better models. In this work, we present an analytic framework to visualize and understand GANs at the unit-, object-, and scene-level. We first identify a group of interpretable units that are closely related to object concepts using a segmentation-based network dissection method. Then, we quantify the causal effect of interpretable units by measuring the ability of interventions to control objects in the output. We examine the contextual relationship between these units and their surroundings by inserting the discovered object concepts into new images. We show several practical applications enabled by our framework, from comparing internal representations across different layers, models, and datasets, to improving GANs by locating and removing artifact-causing units, to interactively manipulating objects in a scene. We provide open source interpretation tools to help researchers and practitioners better understand their GAN models.
https://arxiv.org/abs/1811.10597
Epicormic branches arise from dormant buds patterned during the growth of previous years. Dormant epicormic buds remain on the surface of trees, pushed outward from the pith during secondary growth, but maintaining vascular connections. Epicormic buds can be reactivated, either through natural processes or intentionally, to rejuvenate orchards and control tree architecture. Because epicormic structures are embedded within secondary growth, tomographic approaches are a useful method to study them and understand their development. We apply techniques from image processing to determine the locations of epicormic vascular traces embedded within secondary growth of sweet cherry (Prunus avium L.), revealing the juvenile phyllotactic pattern in the trunk of an adult tree. Techniques include breadth-first search to find the pith of the tree, edge detection to approximate the radius, and a conversion to polar coordinates to threshold and segment phyllotactic features. Intensity values from Magnetic Resonance Imaging (MRI) of the trunk are projected onto the surface of a perfect cylinder to find the locations of traces in the “boundary image”. Mathematical phyllotaxy provides a means to capture the patterns in the boundary image by modeling phyllotactic parameters. Our cherry tree specimen has the conspicuous parastichy pair $(2,3)$, phyllotactic fraction 2/5, and divergence angle of approximately 143 degrees. The methods described not only provide a framework to study phyllotaxy, but for image processing of volumetric image data in plants. Our results have practical implications for orchard rejuvenation and directed approaches to influence tree architecture. The study of epicormic structures, which are hidden within secondary growth, using tomographic methods also opens the possibility of studying the genetic and environmental basis of such structures.
https://arxiv.org/abs/1812.03321
The threat of smuggled nuclear/radiological weapons and material in commercial containerized cargo remains a significant threat to global security more than a decade after the enactment of laws in the United States and elsewhere mandating interdiction efforts. While significant progress has been made towards deploying passive radiation detection systems in maritime ports, such systems are incapable of detecting shielded threats or even certain scenarios in which material is unshielded. Research efforts towards developing systems for detecting such threats have typically focused on the development of systems that are highly-specific to nuclear/radiological threats and no such systems have been widely deployed. While most existing commercially-available cargo radiography systems are not specifically designed for this interdiction task, if items resembling nuclear/radiological threats are sufficiently rare in cargo radiographs to limit false alarms to an acceptably low frequency, then a smuggling interdiction scheme based on existing technology may be feasible. This analysis characterizes the relevant nuclear and radiological threats that may evade detection by passive monitors, and utilizes a dataset of 122,500 stream-of-commerce cargo container images from a 6 MeV endpoint gamma radiography system to determine the frequency at which objects of similar size and density to such threats occur in containers. It is found that for a broad class of threats, including assembled fission devices, gamma radiography is sufficient to flag threats in this cargo stream at false positive rates of $\lesssim$2%.
https://arxiv.org/abs/1901.03753
In this paper, we formulate the color constancy task as an image-to-image translation problem using GANs. By conducting a large set of experiments on different datasets, an experimental survey is provided on the use of different types of GANs to solve for color constancy i.e. CC-GANs (Color Constancy GANs). Based on the experimental review, recommendations are given for the design of CC-GAN architectures based on different criteria, circumstances and datasets.
https://arxiv.org/abs/1812.03085
Neural network architectures found by sophistic search algorithms achieve strikingly good test performance, surpassing most human-crafted network models by significant margins. Although computationally efficient, their design is often very complex, impairing execution speed. Additionally, finding models outside of the search space is not possible by design. While our space is still limited, we implement undiscoverable expert knowledge into the economic search algorithm Efficient Neural Architecture Search (ENAS), guided by the design principles and architecture of ShuffleNet V2. While maintaining baseline-like 2.85% test error on CIFAR-10, our ShuffleNASNets are significantly less complex, require fewer parameters, and are two times faster than the ENAS baseline in a classification task. These models also scale well to a low parameter space, achieving less than 5% test error with little regularization and only 236K parameters.
https://arxiv.org/abs/1812.02975
Classification tasks are usually analysed and improved through new model architectures or hyperparameter optimisation but the underlying properties of datasets are discovered on an ad-hoc basis as errors occur. However, understanding the properties of the data is crucial in perfecting models. In this paper we analyse exactly which characteristics of a dataset best determine how difficult that dataset is for the task of text classification. We then propose an intuitive measure of difficulty for text classification datasets which is simple and fast to calculate. We show that this measure generalises to unseen data by comparing it to state-of-the-art datasets and results. This measure can be used to analyse the precise source of errors in a dataset and allows fast estimation of how difficult a dataset is to learn. We searched for this measure by training 12 classical and neural network based models on 78 real-world datasets, then use a genetic algorithm to discover the best measure of difficulty. Our difficulty-calculating code ( this https URL ) and datasets ( this http URL ) are publicly available.
https://arxiv.org/abs/1811.01910
Automatically generating a natural language sentence to describe the content of an input video is a very challenging problem. It is an essential multimodal task in which auditory and visual contents are equally important. Although audio information has been exploited to improve video captioning in previous works, it is usually regarded as an additional feature fed into a black box fusion machine. How are the words in the generated sentences associated with the auditory and visual modalities? The problem is still not investigated. In this paper, we make the first attempt to design an interpretable audio-visual video captioning network to discover the association between words in sentences and audio-visual sequences. To achieve this, we propose a multimodal convolutional neural network-based audio-visual video captioning framework and introduce a modality-aware module for exploring modality selection during sentence generation. Besides, we collect new audio captioning and visual captioning datasets for further exploring the interactions between auditory and visual modalities for high-level video understanding. Extensive experiments demonstrate that the modality-aware module makes our model interpretable on modality selection during sentence generation. Even with the added interpretability, our video captioning network can still achieve comparable performance with recent state-of-the-art methods.
https://arxiv.org/abs/1812.02872
In this work we propose a new task called Story Visualization. Given a multi-sentence paragraph, the story is visualized by generating a sequence of images, one for each sentence. In contrast to video generation, story visualization focuses less on the continuity in generated images (frames), but more on the global consistency across dynamic scenes and characters – a challenge that has not been addressed by any single-image or video generation methods. Therefore, we propose a new story-to-image-sequence generation model, StoryGAN, based on the sequential conditional GAN framework. Our model is unique in that it consists of a deep Context Encoder that dynamically tracks the story flow, and two discriminators at the story and image levels, respectively, to enhance the image quality and the consistency of the generated sequences. To evaluate the model, we modified existing datasets to create the CLEVR-SV and Pororo-SV datasets. Empirically, StoryGAN outperformed state-of-the-art models in image quality, contextual consistency metrics, and human evaluation.
https://arxiv.org/abs/1812.02784
Time-domain radio astronomy utilizes a harmonic sum algorithm as part of the Fourier domain periodicity search, this type of search is used to discover single pulsars. The harmonic sum algorithm is also used as part of the Fourier domain acceleration search which aims to discover pulsars that are locked in orbit around another pulsar or compact object. However porting the harmonic sum to many-core architectures like GPUs is not a straightforward task. The main problem that must be overcome is the very unfavourable memory access pattern, which gets worse as the dimensionality of the harmonic sum increases. We present a set of algorithms for calculating the harmonic sum that are more suited to many-core architectures such as GPUs. We present an evaluation of the sensitivity of these different approaches, and their performance. This work forms part of the AstroAccelerate project which is a GPU accelerated software package for processing time-domain radio astronomy data.
https://arxiv.org/abs/1812.02647
The extension of image generation to video generation turns out to be a very difficult task, since the temporal dimension of videos introduces an extra challenge during the generation process. Besides, due to the limitation of memory and training stability, the generation becomes increasingly challenging with the increase of the resolution/duration of videos. In this work, we exploit the idea of progressive growing of Generative Adversarial Networks (GANs) for higher resolution video generation. In particular, we begin to produce video samples of low-resolution and short-duration, and then progressively increase both resolution and duration alone (or jointly) by adding new spatiotemporal convolutional layers to the current networks. Starting from the learning on a very raw-level spatial appearance and temporal movement of the video distribution, the proposed progressive method learns spatiotemporal information incrementally to generate higher resolution videos. Furthermore, we introduce a sliced version of Wasserstein GAN (SWGAN) loss to improve the distribution learning on the video data of high-dimension and mixed-spatiotemporal distribution. SWGAN loss replaces the distance between joint distributions by that of one-dimensional marginal distributions, making the loss easier to compute. We evaluate the proposed model on our collected face video dataset of 10,900 videos to generate photorealistic face videos of 256x256x32 resolution. In addition, our model also reaches a record inception score of 14.57 in unsupervised action recognition dataset UCF-101.
https://arxiv.org/abs/1810.02419
Object detection in video is crucial for many applications. Compared to images, video provides additional cues which can help to disambiguate the detection problem. Our goal in this paper is to learn discriminative models for the temporal evolution of object appearance and to use such models for object detection. To model temporal evolution, we introduce space-time tubes corresponding to temporal sequences of bounding boxes. We propose two CNN architectures for generating and classifying tubes, respectively. Our tube proposal network (TPN) first generates a large number of spatio-temporal tube proposals maximizing object recall. The Tube-CNN then implements a tube-level object detector in the video. Our method improves state of the art on two large-scale datasets for object detection in video: HollywoodHeads and ImageNet VID. Tube models show particular advantages in difficult dynamic scenes.
https://arxiv.org/abs/1812.02619
Human motion prediction model has applications in various fields of computer vision. Without taking into account the inherent stochasticity in the prediction of future pose dynamics, such methods often converges to a deterministic undesired mean of multiple probable outcomes. Devoid of this, we propose a novel probabilistic generative approach called Bidirectional Human motion prediction GAN, or BiHMP-GAN. To be able to generate multiple probable human-pose sequences, conditioned on a given starting sequence, we introduce a random extrinsic factor r, drawn from a predefined prior distribution. Furthermore, to enforce a direct content loss on the predicted motion sequence and also to avoid mode-collapse, a novel bidirectional framework is incorporated by modifying the usual discriminator architecture. The discriminator is trained also to regress this extrinsic factor r, which is used alongside with the intrinsic factor (encoded starting pose sequence) to generate a particular pose sequence. To further regularize the training, we introduce a novel recursive prediction strategy. In spite of being in a probabilistic framework, the enhanced discriminator architecture allows predictions of an intermediate part of pose sequence to be used as a conditioning for prediction of the latter part of the sequence. The bidirectional setup also provides a new direction to evaluate the prediction quality against a given test sequence. For a fair assessment of BiHMP-GAN, we report performance of the generated motion sequence using (i) a critic model trained to discriminate between real and fake motion sequence, and (ii) an action classifier trained on real human motion dynamics. Outcomes of both qualitative and quantitative evaluations, on the probabilistic generations of the model, demonstrate the superiority of BiHMP-GAN over previously available methods.
https://arxiv.org/abs/1812.02591
The rapidly increasing interest in nanowires (NWs) of GaN and associated III-Nitrides for (opto-)electronic applications demands immediate addressal of the technological challenges associated with both NW-growth and device processing. Towards this end, we demonstrate in this work an approach to suppress thermal decomposition of GaN NWs, which also serves to remedy the effect of NW-coalescence during growth. While both these effects are well-known to be major hurdles in the development of GaN-NW-devices, reliable methods to tackle these issues have not been reported so far. We show that by providing a thin AlN cap layer, which epitaxially grows only on the top-facet of the GaN NWs, thermal decomposition can be almost completely suppressed. Thermal annealing of GaN NW-ensembles, post AlN-capping, leads to selective decomposition of uncapped/partially-capped NWs, leaving behind (mostly) AlN-capped GaN NWs, with superior crystal- and luminescence characteristics. This simple yet extremely effective approach may therefore serve as a very crucial milestone in the roadmap of GaN-NW-based (opto-)electronic technology.
https://arxiv.org/abs/1812.02443
Distributing Neural Network training is of particular interest for several reasons including scaling using computing clusters, training at data sources such as IOT devices and edge servers, utilizing underutilized resources across heterogeneous environments, and so on. Most contemporary approaches primarily address scaling using computing clusters and require high network bandwidth and frequent communication. This thesis presents an overview of standard approaches to distribute training and proposes a novel technique involving pairwise-communication using Gossip-like protocols, called Elastic Gossip. This approach builds upon an existing technique known as Elastic Averaging SGD (EASGD), and is similar to another technique called Gossiping SGD which also uses Gossip-like protocols. Elastic Gossip is empirically evaluated against Gossiping SGD using the MNIST digit recognition and CIFAR-10 classification tasks, using commonly used Neural Network architectures spanning Multi-Layer Perceptrons (MLPs) and Convolutional Neural Networks (CNNs). It is found that Elastic Gossip, Gossiping SGD, and All-reduce SGD perform quite comparably, even though the latter entails a substantially higher communication cost. While Elastic Gossip performs better than Gossiping SGD in these experiments, it is possible that a more thorough search over hyper-parameter space, specific to a given application, may yield configurations of Gossiping SGD that work better than Elastic Gossip.
https://arxiv.org/abs/1812.02407
The research scope of this paper is to apply spatial object based image analysis (OBIA) method for processing panchromatic multispectral image covering study area of Brussels for urban mapping. The aim is to map different land cover types and more specifically, built-up areas from the very high resolution (VHR) satellite image using OBIA approach. A case study covers urban landscapes in the eastern areas of the city of Brussels, Belgium. Technically, this research was performed in eCognition raster processing software demonstrating excellent results of image segmentation and classification. The tools embedded in eCognition enabled to perform image segmentation and objects classification processes in a semi-automated regime, which is useful for the city planning, spatial analysis and urban growth analysis. The combination of the OBIA method together with technical tools of the eCognition demonstrated applicability of this method for urban mapping in densely populated areas, e.g. in megapolis and capital cities. The methodology included multiresolution segmentation and classification of the created objects.
https://arxiv.org/abs/1901.00726
A detailed environment perception is a crucial component of automated vehicles. However, to deal with the amount of perceived information, we also require segmentation strategies. Based on a grid map environment representation, well-suited for sensor fusion, free-space estimation and machine learning, we detect and classify objects using deep convolutional neural networks. As input for our networks we use a multi-layer grid map efficiently encoding 3D range sensor information. The inference output consists of a list of rotated bounding boxes with associated semantic classes. We conduct extensive ablation studies, highlight important design considerations when using grid maps and evaluate our models on the KITTI Bird’s Eye View benchmark. Qualitative and quantitative benchmark results show that we achieve robust detection and state of the art accuracy solely using top-view grid maps from range sensor data.
https://arxiv.org/abs/1805.08689
This work aims to solve the challenging few-shot object detection problem where only a few annotated examples are available for each object category to train a detection model. Such an ability of learning to detect an object from just a few examples is common for human vision systems, but remains absent for computer vision systems. Though few-shot meta learning offers a promising solution technique, previous works mostly target the task of image classification and are not directly applicable for the much more complicated object detection task. In this work, we propose a novel meta-learning based model with carefully designed architecture, which consists of a meta-model and a base detection model. The base detection model is trained on several base classes with sufficient samples to offer basis features. The meta-model is trained to reweight importance of features from the base detection model over the input image and adapt these features to assist novel object detection from a few examples. The meta-model is light-weight, end-to-end trainable and able to entail the base model with detection ability for novel objects fast. Through experiments we demonstrated our model can outperform baselines by a large margin for few-shot object detection, on multiple datasets and settings. Our model also exhibits fast adaptation speed to novel few-shot classes.
https://arxiv.org/abs/1812.01866
The field of automatic video generation has received a boost thanks to the recent Generative Adversarial Networks (GANs). However, most existing methods cannot control the contents of the generated video using a text caption, losing their usefulness to a large extent. This particularly affects human videos due to their great variety of actions and appearances. This paper presents Conditional Flow and Texture GAN (CFT-GAN), a GAN-based video generation method from action-appearance captions. We propose a novel way of generating video by encoding a caption (e.g., “a man in blue jeans is playing golf”) in a two-stage generation pipeline. Our CFT-GAN uses such caption to generate an optical flow (action) and a texture (appearance) for each frame. As a result, the output video reflects the content specified in the caption in a plausible way. Moreover, to train our method, we constructed a new dataset for human video generation with captions. We evaluated the proposed method qualitatively and quantitatively via an ablation study and a user study. The results demonstrate that CFT-GAN is able to successfully generate videos containing the action and appearances indicated in the captions.
https://arxiv.org/abs/1812.01261
The current- and voltage-scaled sensitivities and signal-to-noise ratios (SNR) (with respect to thermal noise) of various octagonal AlGaN/GaN and InAlN/GaN Hall-effect sensors were examined in this work. The effect of metal contact lengths on sensitivity and sensor offset was evaluated. Calculations that take into account the shape of the device show that devices with point-like contacts have the highest current-scaled sensitivity (68.9 V/A/T), while devices with contacts of equal length to their non-contact sides have the highest voltage-scaled sensitivity (86.9 mV/V/T). The sensitivities of the two other devices follow the predicted trends closely. All the devices have offsets less than 20 $\mu$T at low supply current operation (< 300 $\mu$A) and most remain below 35 $\mu$T at higher supply current (up to 1.2 mA). The consistent low offsets across the devices imply that the choice of Hall-effect sensor geometry should mainly depend on whether the device is current-biased or voltage-biased and the frequency at which it will operate. This work demonstrates that GaN Hall-effect sensor performance can be improved by adjusting the geometry of the Hall-effect plate specific to its function (e.g., power electronics, navigation, automotive applications).
https://arxiv.org/abs/1812.01078
The Emu Chick prototype implements migratory memory-side processing in a novel hardware system. Rather than transferring large amounts of data across the system interconnect, the Emu Chick moves lightweight thread contexts to near-memory cores before the beginning of each remote memory read. Previous work has characterized the performance of the Chick prototype in terms of memory bandwidth and programming differences from more typical, non-migratory platforms, but there has not yet been an analysis of algorithms on this system. This work evaluates irregular algorithms that could benefit from the lightweight, memory-side processing of the Chick and demonstrates techniques and optimization strategies for achieving performance in sparse matrix-vector multiply operation (SpMV), breadth-first search (BFS), and graph alignment across up to eight distributed nodes encompassing 64 nodelets in the Chick system. We also define and justify relative metrics to compare prototype FPGA-based hardware with established ASIC architectures. The Chick currently supports up to 68x scaling for graph alignment, 80 MTEPS for BFS on balanced graphs, and 50\% of measured STREAM bandwidth for SpMV.
https://arxiv.org/abs/1901.02775
It is well known that the generative adversarial nets (GANs) are remarkably difficult to train. The recently proposed Wasserstein GAN (WGAN) creates principled research directions towards addressing these issues. But we found in practice that gradient penalty WGANs (GP-WGANs) still suffer from training instability. In this paper, we combine a Total Variational (TV) regularizing term into the WGAN formulation instead of weight clipping or gradient penalty, which implies that the Lipschitz constraint is enforced on the critic network. Our proposed method is more stable at training than GP-WGANs and works well across varied GAN architectures. We also present a method to control the trade-off between image diversity and visual quality. It does not bring any computation burden.
https://arxiv.org/abs/1812.00810
Recent studies on neural architecture search have shown that automatically designed neural networks perform as good as expert-crafted architectures. While most existing works aim at finding architectures that optimize the prediction accuracy, these architectures may have complexity and is therefore not suitable being deployed on certain computing environment (e.g., with limited power budgets). We propose MONAS, a framework for Multi-Objective Neural Architectural Search that employs reward functions considering both prediction accuracy and other important objectives (e.g., power consumption) when searching for neural network architectures. Experimental results showed that, compared to the state-ofthe-arts, models found by MONAS achieve comparable or better classification accuracy on computer vision applications, while satisfying the additional objectives such as peak power.
https://arxiv.org/abs/1806.10332
It is still challenging to build an AI system that can perform tasks that involve vision and language at human level. So far, researchers have singled out individual tasks separately, for each of which they have designed networks and trained them on its dedicated datasets. Although this approach has seen a certain degree of success, it comes with difficulties of understanding relations among different tasks and transferring the knowledge learned for a task to others. We propose a multi-task learning approach that enables to learn vision-language representation that is shared by many tasks from their diverse datasets. The representation is hierarchical, and prediction for each task is computed from the representation at its corresponding level of the hierarchy. We show through experiments that our method consistently outperforms previous single-task-learning methods on image caption retrieval, visual question answering, and visual grounding. We also analyze the learned hierarchical representation by visualizing attention maps generated in our network.
https://arxiv.org/abs/1812.00500
In this paper, we study the problem of improving computational resource utilization of neural networks. Deep neural networks are usually over-parameterized for their tasks in order to achieve good performances, thus are likely to have underutilized computational resources. This observation motivates a lot of research topics, e.g. network pruning, architecture search, etc. As models with higher computational costs (e.g. more parameters or more computations) usually have better performances, we study the problem of improving the resource utilization of neural networks so that their potentials can be further realized. To this end, we propose a novel optimization method named Neural Rejuvenation. As its name suggests, our method detects dead neurons and computes resource utilization in real time, rejuvenates dead neurons by resource reallocation and reinitialization, and trains them with new training schemes. By simply replacing standard optimizers with Neural Rejuvenation, we are able to improve the performances of neural networks by a very large margin while using similar training efforts and maintaining their original resource usages.
https://arxiv.org/abs/1812.00481
In this paper, we propose a general approach to optimize anchor boxes for object detection. Nowadays, anchor boxes are widely adopted in state-of-the-art detection frameworks. However, all these frameworks pre-define anchor box shapes in a heuristic way and fix the size during training. To improve the accuracy and reduce the effort to design the anchor boxes, we propose to dynamically learn the shapes, which allows the anchors to automatically adapt to the data distribution and the network learning capability. The learning approach can be easily implemented in the stochastic gradient descent way and be plugged into any anchor box-based detection framework. The extra training cost is almost negligible and it has no impact on the inference time cost. Exhaustive experiments also demonstrate that the proposed anchor optimization method consistently achieves significant improvement ($\ge 1\%$ mAP absolute gain) over the baseline method on several benchmark datasets including Pascal VOC 07+12, MS COCO and Brainwash. Meanwhile, the robustness is also verified towards different anchor box initialization methods, which greatly simplifies the problem of anchor box design.
https://arxiv.org/abs/1812.00469
Expectation maximization (EM) algorithm is to find maximum likelihood solution for models having latent variables. A typical example is Gaussian Mixture Model (GMM) which requires Gaussian assumption, however, natural images are highly non-Gaussian so that GMM cannot be applied to perform clustering task on pixel space. To overcome such limitation, we propose a GAN based EM learning framework that can maximize the likelihood of images and estimate the latent variables with only the constraint of L-Lipschitz continuity. We call this model GAN-EM, which is a framework for image clustering, semi-supervised classification and dimensionality reduction. In M-step, we design a novel loss function for discriminator of GAN to perform maximum likelihood estimation (MLE) on data with soft class label assignments. Specifically, a conditional generator captures data distribution for $K$ classes, and a discriminator tells whether a sample is real or fake for each class. Since our model is unsupervised, the class label of real data is regarded as latent variable, which is estimated by an additional network (E-net) in E-step. The proposed GAN-EM achieves state-of-the-art clustering and semi-supervised classification results on MNIST, SVHN and CelebA, as well as comparable quality of generated images to other recently developed generative models.
https://arxiv.org/abs/1812.00335
3D object detection and pose estimation from a single image are two inherently ambiguous problems. Oftentimes, objects appear similar from different viewpoints due to shape symmetries, occlusion and repetitive textures. This ambiguity in both detection and pose estimation means that an object instance can be perfectly described by several different poses and even classes. In this work we propose to explicitly deal with this uncertainty. For each object instance we predict multiple pose and class outcomes to estimate the specific pose distribution generated by symmetries and repetitive textures. The distribution collapses to a single outcome when the visual appearance uniquely identifies just one valid pose. We show the benefits of our approach which provides not only a better explanation for pose ambiguity, but also a higher accuracy in terms of pose estimation.
https://arxiv.org/abs/1812.00287
A key solution to visual question answering (VQA) exists in how to fuse visual and language features extracted from an input image and question. We show that an attention mechanism that enables dense, bi-directional interactions between the two modalities contributes to boost accuracy of prediction of answers. Specifically, we present a simple architecture that is fully symmetric between visual and language representations, in which each question word attends on image regions and each image region attends on question words. It can be stacked to form a hierarchy for multi-step interactions between an image-question pair. We show through experiments that the proposed architecture achieves a new state-of-the-art on VQA and VQA 2.0 despite its small size. We also present qualitative evaluation, demonstrating how the proposed attention mechanism can generate reasonable attention maps on images and questions, which leads to the correct answer prediction.
https://arxiv.org/abs/1804.00775
Object detection in aerial images is an active yet challenging task in computer vision because of the birdview perspective, the highly complex backgrounds, and the variant appearances of objects. Especially when detecting densely packed objects in aerial images, methods relying on horizontal proposals for common object detection often introduce mismatches between the Region of Interests (RoIs) and objects. This leads to the common misalignment between the final object classification confidence and localization accuracy. Although rotated anchors have been used to tackle this problem, the design of them always multiplies the number of anchors and dramatically increases the computational complexity. In this paper, we propose a RoI Transformer to address these problems. More precisely, to improve the quality of region proposals, we first designed a Rotated RoI (RRoI) learner to transform a Horizontal Region of Interest (HRoI) into a Rotated Region of Interest (RRoI). Based on the RRoIs, we then proposed a Rotated Position Sensitive RoI Align (RPS-RoI-Align) module to extract rotation-invariant features from them for boosting subsequent classification and regression. Our RoI Transformer is with light weight and can be easily embedded into detectors for oriented object detection. A simple implementation of the RoI Transformer has achieved state-of-the-art performances on two common and challenging aerial datasets, i.e., DOTA and HRSC2016, with a neglectable reduction to detection speed. Our RoI Transformer exceeds the deformable Position Sensitive RoI pooling when oriented bounding-box annotations are available. Extensive experiments have also validated the flexibility and effectiveness of our RoI Transformer. The results demonstrate that it can be easily integrated with other detector architectures and significantly improve the performances.
https://arxiv.org/abs/1812.00155
The labeling cost of large number of bounding boxes is one of the main challenges for training modern object detectors. To reduce the dependence on expensive bounding box annotations, we propose a new semi-supervised object detection formulation, in which a few seed box level annotations and a large scale of image level annotations are used to train the detector. We adopt a training-mining framework, which is widely used in weakly supervised object detection tasks. However, the mining process inherently introduces various kinds of labelling noises: false negatives, false positives and inaccurate boundaries, which can be harmful for training the standard object detectors (e.g. Faster RCNN). We propose a novel NOise Tolerant Ensemble RCNN (NOTE-RCNN) object detector to handle such noisy labels. Comparing to standard Faster RCNN, it contains three highlights: an ensemble of two classification heads and a distillation head to avoid overfitting on noisy labels and improve the mining precision, masking the negative sample loss in box predictor to avoid the harm of false negative labels, and training box regression head only on seed annotations to eliminate the harm from inaccurate boundaries of mined bounding boxes. We evaluate the methods on ILSVRC 2013 and MSCOCO 2017 dataset; we observe that the detection accuracy consistently improves as we iterate between mining and training steps, and state-of-the-art performance is achieved.
https://arxiv.org/abs/1812.00124
Recent work in network quantization has substantially reduced the time and space complexity of neural network inference, enabling their deployment on embedded and mobile devices with limited computational and memory resources. However, existing quantization methods often represent all weights and activations with the same precision (bit-width). In this paper, we explore a new dimension of the design space: quantizing different layers with different bit-widths. We formulate this problem as a neural architecture search problem and propose a novel differentiable neural architecture search (DNAS) framework to efficiently explore its exponential search space with gradient-based optimization. Experiments show we surpass the state-of-the-art compression of ResNet on CIFAR-10 and ImageNet. Our quantized models with 21.1x smaller model size or 103.9x lower computational cost can still outperform baseline quantized or even full precision models.
https://arxiv.org/abs/1812.00090
Current Visual Question Answering (VQA) systems can answer intelligent questions about Known' visual content. However, their performance drops significantly when questions about visually and linguistically
Unknown’ concepts are presented during inference (Open-world' scenario). A practical VQA system should be able to deal with novel concepts in real world settings. To address this problem, we propose an exemplar-based approach that transfers learning (i.e., knowledge) from previously
Known’ concepts to answer questions about the `Unknown’. We learn a highly discriminative joint embedding space, where visual and semantic features are fused to give a unified representation. Once novel concepts are presented to the model, it looks for the closest match from an exemplar set in the joint embedding space. This auxiliary information is used alongside the given Image-Question pair to refine visual attention in a hierarchical fashion. Since handling the high dimensional exemplars on large datasets can be a significant challenge, we introduce an efficient matching scheme that uses a compact feature description for search and retrieval. To evaluate our model, we propose a new split for VQA, separating Unknown visual and semantic concepts from the training set. Our approach shows significant improvements over state-of-the-art VQA models on the proposed Open-World VQA dataset and standard VQA datasets.
https://arxiv.org/abs/1811.12772
Non-local patch based methods were until recently state-of-the-art for image denoising but are now outperformed by CNNs. Yet they are still the best ones for video denoising, as video redundancy is a key factor to attain high denoising performance. The problem is that CNN architectures are hardly compatible with the search for self-similarities. In this work we propose a new and efficient way to feed video self-similarities to a CNN. The non-locality is incorporated into the network via a first non-trainable layer which finds for each patch in the input image its most similar patches in a search region. The central values of these patches are then gathered in a feature vector which is assigned to each image pixel. This information is presented to a CNN which is trained to predict the clean image. We apply the proposed architecture to image and video denoising. For the latter patches are searched for in a 3D spatio-temporal volume. The proposed architecture achieves state-of-the-art results, specially in video denoising, where it outperforms most state-of-the-art methods. To the best of our knowledge, this is the first successful application of a CNN to video denoising.
https://arxiv.org/abs/1811.12758
SMILES is a linear representation of chemical structures which encodes the connection table, and the stereochemistry of a molecule as a line of text with a grammar structure denoting atoms, bonds, rings and chains, and this information can be used to predict chemical properties. Molecular fingerprints are representations of chemical structures, successfully used in similarity search, clustering, classification, drug discovery, and virtual screening and are a standard and computationally efficient abstract representation where structural features are represented as a bit string. Both SMILES and molecular fingerprints are different representations for describing the structure of a molecule. There exist several predictive models for learning chemical properties based on either SMILES or molecular fingerprints. Here, our goal is to build predictive models that can leverage both these molecular representations. In this work, we present CheMixNet – a set of neural networks for predicting chemical properties from a mixture of features learned from the two molecular representations – SMILES as sequences and molecular fingerprints as vector inputs. We demonstrate the efficacy of CheMixNet architectures by evaluating on six different datasets. The proposed CheMixNet models not only outperforms the candidate neural architectures such as contemporary fully connected networks that uses molecular fingerprints and 1-D CNN and RNN models trained SMILES sequences, but also other state-of-the-art architectures such as Chemception and Molecular Graph Convolutions.
https://arxiv.org/abs/1811.08283
GANs involve training two networks in an adversarial game, where each network’s task depends on its adversary. Recently, several works have framed GAN training as an online or continual learning problem. We focus on the discriminator, which must perform classification under an (adversarially) shifting data distribution. When trained on sequential tasks, neural networks exhibit \emph{forgetting}. For GANs, discriminator forgetting leads to training instability. To counter forgetting, we encourage the discriminator to maintain useful representations by adding a self-supervision. Conditional GANs have a similar effect using labels. However, our self-supervised GAN does not require labels, and closes the performance gap between conditional and unconditional models. We show that, in doing so, the self-supervised discriminator learns better representations than regular GANs.
https://arxiv.org/abs/1810.11598
We propose a principled method for gradient-based regularization of the critic of GAN-like models trained by adversarially optimizing the kernel of a Maximum Mean Discrepancy (MMD). We show that controlling the gradient of the critic is vital to having a sensible loss function, and devise a method to enforce exact, analytical gradient constraints at no additional cost compared to existing approximate techniques based on additive regularizers. The new loss function is provably continuous, and experiments show that it stabilizes and accelerates training, giving image generation models that outperform state-of-the art methods on $160 \times 160$ CelebA and $64 \times 64$ unconditional ImageNet.
https://arxiv.org/abs/1805.11565
Generative adversarial nets (GANs) have generated a lot of excitement. Despite their popularity, they exhibit a number of well-documented issues in practice, which apparently contradict theoretical guarantees. A number of enlightening papers have pointed out that these issues arise from unjustified assumptions that are commonly made, but the message seems to have been lost amid the optimism of recent years. We believe the identified problems deserve more attention, and highlight the implications on both the properties of GANs and the trajectory of research on probabilistic models. We recently proposed an alternative method that sidesteps these problems.
https://arxiv.org/abs/1811.12402
We study in this paper the problems of both image captioning and text-to-image generation, and present a novel turbo learning approach to jointly training an image-to-text generator (a.k.a. CaptionBot) and a text-to-image generator (a.k.a. DrawingBot). The key idea behind the joint training is that image-to-text generation and text-to-image generation as dual problems can form a closed loop to provide informative feedback to each other. Based on such feedback, we introduce a new loss metric by comparing the original input with the output produced by the closed loop. In addition to the old loss metrics used in CaptionBot and DrawingBot, this extra loss metric makes the jointly trained CaptionBot and DrawingBot better than the separately trained CaptionBot and DrawingBot. Furthermore, the turbo-learning approach enables semi-supervised learning since the closed loop can provide pseudo-labels for unlabeled samples. Experimental results on the COCO dataset demonstrate that the proposed turbo learning can significantly improve the performance of both CaptionBot and DrawingBot by a large margin.
https://arxiv.org/abs/1805.08170
An image is not just a collection of objects, but rather a graph where each object is related to other objects through spatial and semantic relations. Using relational reasoning modules, allowing message passing between objects, can therefore improve object detection. Current schemes apply such dedicated modules either on a specific layer of the bottom-up stream, or between already-detected objects. We show that the relational process can be better modeled in a coarse to fine manner and present a novel framework, applying a non-local module sequentially to increasing resolution feature-maps along the top-down stream. In this way, the inner relational process can naturally pass information from larger objects to smaller related ones. Applying the modules to fine feature-maps also allows message passing between the small objects themselves, exploiting repetitions of instances from of the same class. In practice, due to the expensive memory utilization of the non-local module, it is unfeasible to apply the module as currently used to high-resolution feature-maps. We efficiently redesigned the non local module, improved it in terms of memory and number of operations, allowing it to be placed anywhere along the network. We also incorporated relative spatial information into the module, in a manner that can be incorporated into our efficient implementation. We show the effectiveness of our scheme by improving the results of detecting small objects on COCO by 1.5 AP over Faster RCNN and by 1 AP over using non-local module on the bottom-up stream.
https://arxiv.org/abs/1811.12152
Embedded deep learning platforms have witnessed two simultaneous improvements. First, the accuracy of convolutional neural networks (CNNs) has been significantly improved through the use of automated neural-architecture search (NAS) algorithms to determine CNN structure. Second, there has been increasing interest in developing application-specific platforms for CNNs that provide improved inference performance and energy consumption as compared to GPUs. Embedded deep learning platforms differ in the amount of compute resources and memory-access bandwidth, which would affect performance and energy consumption of CNNs. It is therefore critical to consider the available hardware resources in the network architecture search. To this end, we introduce TEA-DNN, a NAS algorithm targeting multi-objective optimization of execution time, energy consumption, and classification accuracy of CNN workloads on embedded architectures. TEA-DNN leverages energy and execution time measurements on embedded hardware when exploring the Pareto-optimal curves across accuracy, execution time, and energy consumption and does not require additional effort to model the underlying hardware. We apply TEA-DNN for image classification on actual embedded platforms (NVIDIA Jetson TX2 and Intel Movidius Neural Compute Stick). We highlight the Pareto-optimal operating points that emphasize the necessity to explicitly consider hardware characteristics in the search process. To the best of our knowledge, this is the most comprehensive study of Pareto-optimal models across a range of hardware platforms using actual measurements on hardware to obtain objective values.
https://arxiv.org/abs/1811.12065
Convolutional Neural Networks (CNNs) usually use the same activation function, such as RELU, for all convolutional layers. There are performance limitations of just using RELU. In order to achieve better classification performance, reduce training and testing times, and reduce power consumption and memory usage, a new “Compressed Multi-function CNN” is developed. Google’s Inception-V4, for example, is a very deep CNN that consists of 4 Inception-A blocks, 7 Inception-B blocks, and 3 Inception-C blocks. RELU is used for all convolutional layers. A new “Compressed Multi-function Inception-V4” (CMI) that can use different activation functions is created with k Inception-A blocks, m Inception-B blocks, and n Inception-C blocks where k in {1, 2, 3, 4}, m in {1, 2, 3, 4, 5, 6, 7}, n in {1, 2, 3}, and (k+m+n)<14. For performance analysis, a dataset for classifying brain MRI images into one of the four stages of Alzheimer’s disease is used to compare three CMI architectures with Inception-V4 in terms of F1-score, training and testing times (related to power consumption), and memory usage (model size). Overall, simulations show that the new CMI models can outperform both the commonly used Inception-V4 and Inception-V4 using different activation functions. In the future, other “Compressed Multi-function CNNs”, such as “Compressed Multi-function ResNets and DenseNets” that have a reduced number of convolutional blocks using different activation functions, will be developed to further increase classification accuracy, reduce training and testing times, reduce computational power, and reduce memory usage (model size) for building more effective healthcare systems, such as implementing accurate and convenient disease diagnosis systems on mobile devices that have limited battery power and memory.
https://arxiv.org/abs/1811.11996
We describe a DNN for video classification and captioning, trained end-to-end, with shared features, to solve tasks at different levels of granularity, exploring the link between granularity in a source task and the quality of learned features for transfer learning. For solving the new task domain in transfer learning, we freeze the trained encoder and fine-tune a neural net on the target domain. We train on the Something-Something dataset with over 220, 000 videos, and multiple levels of target granularity, including 50 action groups, 174 fine-grained action categories and captions. Classification and captioning with Something-Something are challenging because of the subtle differences between actions, applied to thousands of different object classes, and the diversity of captions penned by crowd actors. Our model performs better than existing classification baselines for SomethingSomething, with impressive fine-grained results. And it yields a strong baseline on the new Something-Something captioning task. Experiments reveal that training with more fine-grained tasks tends to produce better features for transfer learning.
https://arxiv.org/abs/1804.09235
In evidence-based medicine (EBM), structured medical questions are always favored for efficient search of the best available evidence for treatments. PICO element detection is widely used to help structurize the clinical studies and question by identifying the sentences in a given medical text that belong to one of the four components: Participants (P), Intervention (I), Comparison (C), and Outcome (O). In this work, we propose a hierarchical deep neural network (DNN) architecture that contains dual bi-directional long short-term memory (bi-LSTM) layers to automatically detect the PICO element in medical texts. Within the model, the lower layer of bi-LSTM is for sentence encoding while the upper one is to contextualize the encoded sentence representation vector. In addition, we adopt adversarial and virtual adversarial training to regularize the model. Overall, we advance the PICO element detection to new state-of-the-art performance, outperforming the previous works by at least 4\% in F1 score for all P/I/O categories.
https://arxiv.org/abs/1810.12780