Introduction of spectrum-sharing in 5G and subsequent generation networks demand base-station(s) with the capability to characterize the wideband spectrum spanned over licensed, shared and unlicensed non-contiguous frequency bands. Spectrum characterization involves the identification of vacant bands along with center frequency and parameters (energy, modulation, etc.) of occupied bands. Such characterization at Nyquist sampling is area and power-hungry due to the need for high-speed digitization. Though sub-Nyquist sampling (SNS) offers an excellent alternative when the spectrum is sparse, it suffers from poor performance at low signal to noise ratio (SNR) and demands careful design and integration of digital reconstruction, tunable channelizer and characterization algorithms. In this paper, we propose a novel deep-learning framework via a single unified pipeline to accomplish two tasks: 1)~Reconstruct the signal directly from sub-Nyquist samples, and 2)~Wideband spectrum characterization. The proposed approach eliminates the need for complex signal conditioning between reconstruction and characterization and does not need complex tunable channelizers. We extensively compare the performance of our framework for a wide range of modulation schemes, SNR and channel conditions. We show that the proposed framework outperforms existing SNS based approaches and characterization performance approaches to Nyquist sampling-based framework with an increase in SNR. Easy to design and integrate along with a single unified deep learning framework make the proposed architecture a good candidate for reconfigurable platforms.

Histopathology is a reflection of the molecular changes and provides prognostic phenotypes representing the disease progression. In this study, we introduced feature scores generated from hematoxylin and eosin histology images based on deep learning (DL) models developed for prostate pathology. We demonstrated that these feature scores were significantly prognostic for time to event endpoints (biochemical recurrence and cancer-specific survival) and had simultaneously molecular biologic associations to relevant genomic alterations and molecular subtypes using already trained DL models that were not previously exposed to the datasets of the current study. Further, we discussed the potential of such feature scores to improve the current tumor grading system and the challenges that are associated with tumor heterogeneity and the development of prognostic models from histology images. Our findings uncover the potential of feature scores from histology images as digital biomarkers in precision medicine and as an expanding utility for digital pathology.

Robots are required to operate autonomously in response to changing situations. Previously, imitation learning using 4ch-bilateral control was demonstrated to be suitable for imitation of object manipulation. However, cooperative work between humans and robots has not yet been verified in these studies. In this study, the task was expanded by cooperative work between a human and a robot. 4ch-bilateral control was used to collect training data for training robot motion. We focused on serving salad as a task in the home. The task was executed with a spoon and a fork fixed to robots. Adjustment of force was indispensable in manipulating indefinitely shaped objects such as salad. Results confirmed the effectiveness of the proposed method as demonstrated by the success of the task.

For real-time semantic video segmentation, most recent works utilised a dynamic framework with a key scheduler to make online key/non-key decisions. Some works used a fixed key scheduling policy, while others proposed adaptive key scheduling methods based on heuristic strategies, both of which may lead to suboptimal global performance. To overcome this limitation, we model the online key decision process in dynamic video segmentation as a deep reinforcement learning problem and learn an efficient and effective scheduling policy from expert information about decision history and from the process of maximising global return. Moreover, we study the application of dynamic video segmentation on face videos, a field that has not been investigated before. By evaluating on the 300VW dataset, we show that the performance of our reinforcement key scheduler outperforms that of various baselines in terms of both effective key selections and running speed. Further results on the Cityscapes dataset demonstrate that our proposed method can also generalise to other scenarios. To the best of our knowledge, this is the first work to use reinforcement learning for online key-frame decision in dynamic video segmentation, and also the first work on its application on face videos.

The Hierarchical Equal Area isoLatitude Pixelation (HEALPix) scheme is used extensively in astrophysics for data collection and analysis on the sphere. The scheme was originally designed for studying the Cosmic Microwave Background (CMB) radiation, which represents the first light to travel during the early stages of the universe's development and gives the strongest evidence for the Big Bang theory to date. Refined analysis of the CMB angular power spectrum can lead to revolutionary developments in understanding the nature of dark matter and dark energy. In this paper, we present a new method for performing spherical harmonic analysis for HEALPix data, which is a central component to computing and analyzing the angular power spectrum of the massive CMB data sets. The method uses a novel combination of a non-uniform fast Fourier transform, the double Fourier sphere method, and Slevinsky's fast spherical harmonic transform (Slevinsky, 2019). For a HEALPix grid with $N$ pixels (points), the computational complexity of the method is $mathcal{O}(Nlog^2 N)$, with an initial set-up cost of $mathcal{O}(N^{3/2}log N)$. This compares favorably with $mathcal{O}(N^{3/2})$ runtime complexity of the current methods available in the HEALPix software when multiple maps need to be analyzed at the same time. Using numerical experiments, we demonstrate that the new method also appears to provide better accuracy over the entire angular power spectrum of synthetic data when compared to the current methods, with a convergence rate at least two times higher.

Finding a good regularization parameter for Tikhonov regularization problems is a though yet often asked question. One approach is to use leave-one-out cross-validation scores to indicate the goodness of fit. This utilizes only the noisy function values but, on the downside, comes with a high computational cost. In this paper we present a general approach to shift the main computations from the function in question to the node distribution and, making use of FFT and FFT-like algorithms, even reduce this cost tremendously to the cost of the Tikhonov regularization problem itself. We apply this technique in different settings on the torus, the unit interval, and the two-dimensional sphere. Given that the sampling points satisfy a quadrature rule our algorithm computes the cross-validations scores in floating-point precision. In the cases of arbitrarily scattered nodes we propose an approximating algorithm with the same complexity. Numerical experiments indicate the applicability of our algorithms.

The objective of deep metric learning (DML) is to learn embeddings that can capture semantic similarity and dissimilarity information among data points. Existing pairwise or tripletwise loss functions used in DML are known to suffer from slow convergence due to a large proportion of trivial pairs or triplets as the model improves. To improve this, ranking-motivated structured losses are proposed recently to incorporate multiple examples and exploit the structured information among them. They converge faster and achieve state-of-the-art performance. In this work, we unveil two limitations of existing ranking-motivated structured losses and propose a novel ranked list loss to solve both of them. First, given a query, only a fraction of data points is incorporated to build the similarity structure. To address this, we propose to build a set-based similarity structure by exploiting all instances in the gallery. The learning setting can be interpreted as few-shot retrieval: given a mini-batch, every example is iteratively used as a query, and the rest ones compose the galley to search, i.e., the support set in few-shot setting. The rest examples are split into a positive set and a negative set. For every mini-batch, the learning objective of ranked list loss is to make the query closer to the positive set than to the negative set by a margin. Second, previous methods aim to pull positive pairs as close as possible in the embedding space. As a result, the intraclass data distribution tends to be extremely compressed. In contrast, we propose to learn a hypersphere for each class in order to preserve useful similarity structure inside it, which functions as regularisation. Extensive experiments demonstrate the superiority of our proposal by comparing with the state-of-the-art methods on the fine-grained image retrieval task.

In this paper we revisit the deterministic version of the Sparse Fourier Transform problem, which asks to read only a few entries of $x in mathbb{C}^n$ and design a recovery algorithm such that the output of the algorithm approximates $hat x$, the Discrete Fourier Transform (DFT) of $x$. The randomized case has been well-understood, while the main work in the deterministic case is that of Merhi et al.@ (J Fourier Anal Appl 2018), which obtains $O(k^2 log^{-1}k cdot log^{5.5}n)$ samples and a similar runtime with the $ell_2/ell_1$ guarantee. We focus on the stronger $ell_{infty}/ell_1$ guarantee and the closely related problem of incoherent matrices. We list our contributions as follows.

1. We find a deterministic collection of $O(k^2 log n)$ samples for the $ell_infty/ell_1$ recovery in time $O(nk log^2 n)$, and a deterministic collection of $O(k^2 log^2 n)$ samples for the $ell_infty/ell_1$ sparse recovery in time $O(k^2 log^3n)$.

2. We give new deterministic constructions of incoherent matrices that are row-sampled submatrices of the DFT matrix, via a derandomization of Bernstein's inequality and bounds on exponential sums considered in analytic number theory. Our first construction matches a previous randomized construction of Nelson, Nguyen and Woodruff (RANDOM'12), where there was no constraint on the form of the incoherent matrix.

Our algorithms are nearly sample-optimal, since a lower bound of $Omega(k^2 + k log n)$ is known, even for the case where the sensing matrix can be arbitrarily designed. A similar lower bound of $Omega(k^2 log n/ log k)$ is known for incoherent matrices.

We develop a supervised machine learning algorithm that is able to learn topological phases for finite condensed matter systems from bulk data in real lattice space. The algorithm employs diagonalization in real space together with any supervised learning algorithm to learn topological phases through an eigenvector ensembling procedure. We combine our algorithm with decision trees and random forests to successfully recover topological phase diagrams of Su-Schrieffer-Heeger (SSH) models from bulk lattice data in real space and show how the Shannon information entropy of ensembles of lattice eigenvectors can be used to retrieve a signal detailing how topological information is distributed in the bulk. The discovery of Shannon information entropy signals associated with topological phase transitions from the analysis of data from several thousand SSH systems illustrates how model explainability in machine learning can advance the research of exotic quantum materials with properties that may power future technological applications such as qubit engineering for quantum computing.

We develop a Learning Direct Optimization (LiDO) method for the refinement of a latent variable model that describes input image x. Our goal is to explain a single image x with an interpretable 3D computer graphics model having scene graph latent variables z (such as object appearance, camera position). Given a current estimate of z we can render a prediction of the image g(z), which can be compared to the image x. The standard way to proceed is then to measure the error E(x, g(z)) between the two, and use an optimizer to minimize the error. However, it is unknown which error measure E would be most effective for simultaneously addressing issues such as misaligned objects, occlusions, textures, etc. In contrast, the LiDO approach trains a Prediction Network to predict an update directly to correct z, rather than minimizing the error with respect to z. Experiments show that our LiDO method converges rapidly as it does not need to perform a search on the error landscape, produces better solutions than error-based competitors, and is able to handle the mismatch between the data and the fitted scene model. We apply LiDO to a realistic synthetic dataset, and show that the method also transfers to work well with real images.

We investigate sparse matrix bipartitioning -- a problem where we minimize the communication volume in parallel sparse matrix-vector multiplication. We prove, by reduction from graph bisection, that this problem is $mathcal{NP}$-complete in the case where each side of the bipartitioning must contain a linear fraction of the nonzeros.

We present an improved exact branch-and-bound algorithm which finds the minimum communication volume for a given matrix and maximum allowed imbalance. The algorithm is based on a maximum-flow bound and a packing bound, which extend previous matching and packing bounds.

We implemented the algorithm in a new program called MP (Matrix Partitioner), which solved 839 matrices from the SuiteSparse collection to optimality, each within 24 hours of CPU-time. Furthermore, MP solved the difficult problem of the matrix cage6 in about 3 days. The new program is on average more than ten times faster than the previous program MondriaanOpt.

Benchmark results using the set of 839 optimally solved matrices show that combining the medium-grain/iterative refinement methods of the Mondriaan package with the hypergraph bipartitioner of the PaToH package produces sparse matrix bipartitionings on average within 10% of the optimal solution.

In this paper, we focus on the task of small-footprint keyword spotting under the far-field scenario. Far-field environments are commonly encountered in real-life speech applications, and it causes serve degradation of performance due to room reverberation and various kinds of noises. Our baseline system is built on the convolutional neural network trained with pooled data of both far-field and close-talking speech. To cope with the distortions, we adopt the multi-task learning scheme with alignment loss to reduce the mismatch between the embedding features learned from different domains of data. Experimental results show that our proposed method maintains the performance on close-talking speech and achieves significant improvement on the far-field test set.

We perform an effectivization of classical results concerning universal coding and prediction for stationary ergodic processes over an arbitrary finite alphabet. That is, we lift the well-known almost sure statements to statements about Martin-L"of random sequences. Most of this work is quite mechanical but, by the way, we complete a result of Ryabko from 2008 by showing that each universal probability measure in the sense of universal coding induces a universal predictor in the prequential sense. Surprisingly, the effectivization of this implication holds true provided the universal measure does not ascribe too low conditional probabilities to individual symbols. As an example, we show that the Prediction by Partial Matching (PPM) measure satisfies this requirement. In the almost sure setting, the requirement is superfluous.

Showing items that do not match search query intent degrades customer experience in e-commerce. These mismatches result from counterfactual biases of the ranking algorithms toward noisy behavioral signals such as clicks and purchases in the search logs. Mitigating the problem requires a large labeled dataset, which is expensive and time-consuming to obtain. In this paper, we develop a deep, end-to-end model that learns to effectively classify mismatches and to generate hard mismatched examples to improve the classifier. We train the model end-to-end by introducing a latent variable into the cross-entropy loss that alternates between using the real and generated samples. This not only makes the classifier more robust but also boosts the overall ranking performance. Our model achieves a relative gain compared to baselines by over 26% in F-score, and over 17% in Area Under PR curve. On live search traffic, our model gains significant improvement in multiple countries.

In ASPIC-style structured argumentation an argument can rebut another argument by attacking its conclusion. Two ways of formalizing rebuttal have been proposed: In restricted rebuttal, the attacked conclusion must have been arrived at with a defeasible rule, whereas in unrestricted rebuttal, it may have been arrived at with a strict rule, as long as at least one of the antecedents of this strict rule was already defeasible. One systematic way of choosing between various possible definitions of a framework for structured argumentation is to study what rationality postulates are satisfied by which definition, for example whether the closure postulate holds, i.e. whether the accepted conclusions are closed under strict rules. While having some benefits, the proposal to use unrestricted rebuttal faces the problem that the closure postulate only holds for the grounded semantics but fails when other argumentation semantics are applied, whereas with restricted rebuttal the closure postulate always holds. In this paper we propose that ASPIC-style argumentation can benefit from keeping track not only of the attack relation between arguments, but also the relation of deductive joint support that holds between a set of arguments and an argument that was constructed from that set using a strict rule. By taking this deductive joint support relation into account while determining the extensions, the closure postulate holds with unrestricted rebuttal under all admissibility-based semantics. We define the semantics of deductive joint support through the flattening method.

Passive optical networks (PON) can play a vital role in data centres and access fog solutions by providing scalable, cost and energy efficient architectures. This paper proposes a Mixed Integer Linear Programming (MILP) model to optimize the placement of virtual machines (VMs) over an energy efficient WDM-TDM AWGR PON based data centre architecture. In this optimization, the use of VMs and their requirements affect the optimum number of servers utilized in the data centre when minimizing the power consumption and enabling more efficient utilization of servers is considered. Two power consumption minimization objectives were examined for up to 20 VMs with different computing and networking requirements. The results indicate that considering the minimization of the processing and networking power consumption in the allocation of VMs in the WDM-TDM AWGR PON can reduce the networking power consumption by up to 70% compared to the minimization of the processing power consumption.

Generative feature matching network (GFMN) is an approach for training implicit generative models for images by performing moment matching on features from pre-trained neural networks. In this paper, we present new GFMN formulations that are effective for sequential data. Our experimental results show the effectiveness of the proposed method, SeqGFMN, for three distinct generation tasks in English: unconditional text generation, class-conditional text generation, and unsupervised text style transfer. SeqGFMN is stable to train and outperforms various adversarial approaches for text generation and text style transfer.

Deep learning-based object detection and instance segmentation have achieved unprecedented progress. In this paper, we propose Complete-IoU (CIoU) loss and Cluster-NMS for enhancing geometric factors in both bounding box regression and Non-Maximum Suppression (NMS), leading to notable gains of average precision (AP) and average recall (AR), without the sacrifice of inference efficiency. In particular, we consider three geometric factors, i.e., overlap area, normalized central point distance and aspect ratio, which are crucial for measuring bounding box regression in object detection and instance segmentation. The three geometric factors are then incorporated into CIoU loss for better distinguishing difficult regression cases. The training of deep models using CIoU loss results in consistent AP and AR improvements in comparison to widely adopted $ell_n$-norm loss and IoU-based loss. Furthermore, we propose Cluster-NMS, where NMS during inference is done by implicitly clustering detected boxes and usually requires less iterations. Cluster-NMS is very efficient due to its pure GPU implementation, , and geometric factors can be incorporated to improve both AP and AR. In the experiments, CIoU loss and Cluster-NMS have been applied to state-of-the-art instance segmentation (e.g., YOLACT), and object detection (e.g., YOLO v3, SSD and Faster R-CNN) models. Taking YOLACT on MS COCO as an example, our method achieves performance gains as +1.7 AP and +6.2 AR$_{100}$ for object detection, and +0.9 AP and +3.5 AR$_{100}$ for instance segmentation, with 27.1 FPS on one NVIDIA GTX 1080Ti GPU. All the source code and trained models are available at https://github.com/Zzh-tju/CIoU

To implement a blockchain, we need a blockchain protocol for all the nodes to follow. To design a blockchain protocol, we need a block publisher selection mechanism and a chain selection rule. In Proof-of-Stake (PoS) based blockchain protocols, block publisher selection mechanism selects the node to publish the next block based on the relative stake held by the node. However, PoS protocols may face vulnerability to fully adaptive corruptions. In literature, researchers address this issue at the cost of performance.

In this paper, we propose a novel PoS-based blockchain protocol, QuickSync, to achieve security against fully adaptive corruptions without compromising on performance. We propose a metric called block power, a value defined for each block, derived from the output of the verifiable random function based on the digital signature of the block publisher. With this metric, we compute chain power, the sum of block powers of all the blocks comprising the chain, for all the valid chains. These metrics are a function of the block publisher's stake to enable the PoS aspect of the protocol. The chain selection rule selects the chain with the highest chain power as the one to extend. This chain selection rule hence determines the selected block publisher of the previous block. When we use metrics to define the chain selection rule, it may lead to vulnerabilities against Sybil attacks. QuickSync uses a Sybil attack resistant function implemented using histogram matching. We prove that QuickSync satisfies common prefix, chain growth, and chain quality properties and hence it is secure. We also show that it is resilient to different types of adversarial attack strategies. Our analysis demonstrates that QuickSync performs better than Bitcoin by an order of magnitude on both transactions per second and time to finality, and better than Ouroboros v1 by a factor of three on time to finality.

Danish is a North Germanic/Scandinavian language spoken primarily in Denmark, a country with a tradition of technological and scientific innovation. However, from a technological perspective, the Danish language has received relatively little attention and, as a result, Danish language technology is hard to develop, in part due to a lack of large or broad-coverage Danish corpora. This paper describes the Danish Gigaword project, which aims to construct a freely-available one billion word corpus of Danish text that represents the breadth of the written language.

We present a new mouse cursor designed to facilitate the use of the mouse by people with peripheral vision loss. The pointer consists of a collection of converging straight lines covering the whole screen and following the position of the mouse cursor. We measured its positive effects with a group of participants with peripheral vision loss of different kinds and we found that it can reduce by a factor of 7 the time required to complete a targeting task using the mouse. Using eye tracking, we show that this system makes it possible to initiate the movement towards the target without having to precisely locate the mouse pointer. Using Fitts' Law, we compare these performances with those of full visual field users in order to understand the relation between the accuracy of the estimated mouse cursor position and the index of performance obtained with our tool.

The history of text can be traced back over thousands of years. Rich and precise semantic information carried by text is important in a wide range of vision-based application scenarios. Therefore, text recognition in natural scenes has been an active research field in computer vision and pattern recognition. In recent years, with the rise and development of deep learning, numerous methods have shown promising in terms of innovation, practicality, and efficiency. This paper aims to (1) summarize the fundamental problems and the state-of-the-art associated with scene text recognition; (2) introduce new insights and ideas; (3) provide a comprehensive review of publicly available resources; (4) point out directions for future work. In summary, this literature review attempts to present the entire picture of the field of scene text recognition. It provides a comprehensive reference for people entering this field, and could be helpful to inspire future research. Related resources are available at our Github repository: https://github.com/HCIILAB/Scene-Text-Recognition.

Unsupervised learning of hidden representations has been one of the most vibrant research directions in machine learning in recent years. In this work we study the brain-like Bayesian Confidence Propagating Neural Network (BCPNN) model, recently extended to extract sparse distributed high-dimensional representations. The saliency and separability of the hidden representations when trained on MNIST dataset is studied using an external classifier, and compared with other unsupervised learning methods that include restricted Boltzmann machines and autoencoders.

Consider the following prediction problem. Assume that there is a block box that produces bits according to some unknown computable distribution on the binary tree. We know first $n$ bits $x_1 x_2 ldots x_n$. We want to know the probability of the event that that the next bit is equal to $1$. Solomonoff suggested to use universal semimeasure $m$ for solving this task. He proved that for every computable distribution $P$ and for every $b in {0,1}$ the following holds: $$sum_{n=1}^{infty}sum_{x: l(x)=n} P(x) (P(b | x) - m(b | x))^2 < infty .$$ However, Solomonoff's method has a negative aspect: Hutter and Muchnik proved that there are an universal semimeasure $m$, computable distribution $P$ and a random (in Martin-L{"o}f sense) sequence $x_1 x_2ldots$ such that $lim_{n o infty} P(x_{n+1} | x_1ldots x_n) - m(x_{n+1} | x_1ldots x_n) rightarrow 0$. We suggest a new way for prediction. For every finite string $x$ we predict the new bit according to the best (in some sence) distribution for $x$. We prove the similar result as Solomonoff theorem for our way of prediction. Also we show that our method of prediction has no that negative aspect as Solomonoff's method.

We study the language universality problem for One-Counter Nets, also known as 1-dimensional Vector Addition Systems with States (1-VASS), parameterized either with an initial counter value, or with an upper bound on the allowed counter value during runs. The language accepted by an OCN (defined by reaching a final control state) is monotone in both parameters. This yields two natural questions: 1) Does there exist an initial counter value that makes the language universal? 2) Does there exist a sufficiently high ceiling so that the bounded language is universal? Despite the fact that unparameterized universality is Ackermann-complete and that these problems seem to reduce to checking basic structural properties of the underlying automaton, we show that in fact both problems are undecidable. We also look into the complexities of the problems for several decidable subclasses, namely for unambiguous, and deterministic systems, and for those over a single-letter alphabet.

Simplified Chinese to Traditional Chinese character conversion is a common preprocessing step in Chinese NLP. Despite this, current approaches have poor performance because they do not take into account that a simplified Chinese character can correspond to multiple traditional characters. Here, we propose a model that can disambiguate between mappings and convert between the two scripts. The model is based on subword segmentation, two language models, as well as a method for mapping between subword sequences. We further construct benchmark datasets for topic classification and script conversion. Our proposed method outperforms previous Chinese Character conversion approaches by 6 points in accuracy. These results are further confirmed in a downstream application, where 2kenize is used to convert pretraining dataset for topic classification. An error analysis reveals that our method's particular strengths are in dealing with code-mixing and named entities.

MineRL 2019 competition challenged participants to train sample-efficient agents to play Minecraft, by using a dataset of human gameplay and a limit number of steps the environment. We approached this task with behavioural cloning by predicting what actions human players would take, and reached fifth place in the final ranking. Despite being a simple algorithm, we observed the performance of such an approach can vary significantly, based on when the training is stopped. In this paper, we detail our submission to the competition, run further experiments to study how performance varied over training and study how different engineering decisions affected these results.

Wireless sensor networks have attracted much attention because of many applications in the fields of industry, military, medicine, agriculture, and education. In addition, the vast majority of researches has been done to expand its applications and improve its efficiency. However, there are still many challenges for increasing the efficiency in different parts of this network. One of the most important parts is to improve the network lifetime in the wireless sensor network. Since the sensor nodes are generally powered by batteries, the most important issue to consider in these types of networks is to reduce the power consumption of the nodes in such a way as to increase the network lifetime to an acceptable level. The contribution of this paper is using topology control, the threshold for the remaining energy in nodes, and two of the meta-algorithms include SA (Simulated annealing) and VNS (Variable Neighbourhood Search) to increase the energy remaining in the sensors. Moreover, using a low-cost spanning tree, an appropriate connectivity control among nodes is created in the network in order to increase the network lifetime. The results of simulations show that the proposed method improves the sensor lifetime and reduces the energy consumed.

Neural machine translation (NMT) needs large parallel corpora for state-of-the-art translation quality. Low-resource NMT is typically addressed by transfer learning which leverages large monolingual or parallel corpora for pre-training. Monolingual pre-training approaches such as MASS (MAsked Sequence to Sequence) are extremely effective in boosting NMT quality for languages with small parallel corpora. However, they do not account for linguistic information obtained using syntactic analyzers which is known to be invaluable for several Natural Language Processing (NLP) tasks. To this end, we propose JASS, Japanese-specific Sequence to Sequence, as a novel pre-training alternative to MASS for NMT involving Japanese as the source or target language. JASS is joint BMASS (Bunsetsu MASS) and BRSS (Bunsetsu Reordering Sequence to Sequence) pre-training which focuses on Japanese linguistic units called bunsetsus. In our experiments on ASPEC Japanese--English and News Commentary Japanese--Russian translation we show that JASS can give results that are competitive with if not better than those given by MASS. Furthermore, we show for the first time that joint MASS and JASS pre-training gives results that significantly surpass the individual methods indicating their complementary nature. We will release our code, pre-trained models and bunsetsu annotated data as resources for researchers to use in their own NLP tasks.

Hypertension is a potentially unsafe health ailment, which can be indicated directly from the Blood pressure (BP). Hypertension always leads to other health complications. Continuous monitoring of BP is very important; however, cuff-based BP measurements are discrete and uncomfortable to the user. To address this need, a cuff-less, continuous and a non-invasive BP measurement system is proposed using Photoplethysmogram (PPG) signal and demographic features using machine learning (ML) algorithms. PPG signals were acquired from 219 subjects, which undergo pre-processing and feature extraction steps. Time, frequency and time-frequency domain features were extracted from the PPG and their derivative signals. Feature selection techniques were used to reduce the computational complexity and to decrease the chance of over-fitting the ML algorithms. The features were then used to train and evaluate ML algorithms. The best regression models were selected for Systolic BP (SBP) and Diastolic BP (DBP) estimation individually. Gaussian Process Regression (GPR) along with ReliefF feature selection algorithm outperforms other algorithms in estimating SBP and DBP with a root-mean-square error (RMSE) of 6.74 and 3.59 respectively. This ML model can be implemented in hardware systems to continuously monitor BP and avoid any critical health conditions due to sudden changes.

Recent works imply that the channel pruning can be regarded as searching optimal sub-structure from unpruned networks.

However, existing works based on this observation require training and evaluating a large number of structures, which limits their application.

In this paper, we propose a novel differentiable method for channel pruning, named Differentiable Markov Channel Pruning (DMCP), to efficiently search the optimal sub-structure.

Our method is differentiable and can be directly optimized by gradient descent with respect to standard task loss and budget regularization (e.g. FLOPs constraint).

In DMCP, we model the channel pruning as a Markov process, in which each state represents for retaining the corresponding channel during pruning, and transitions between states denote the pruning process.

In the end, our method is able to implicitly select the proper number of channels in each layer by the Markov process with optimized transitions. To validate the effectiveness of our method, we perform extensive experiments on Imagenet with ResNet and MobilenetV2.

Results show our method can achieve consistent improvement than state-of-the-art pruning methods in various FLOPs settings. The code is available at https://github.com/zx55/dmcp

The paper develops a robotic manipulation system to treat the pressing needs for handling a large number of test tubes in clinical examination and replace or reduce human labor. It presents the technical details of the system, which separates and arranges test tubes in racks with the help of 3D vision and artificial intelligence (AI) reasoning/planning. The developed system only requires a person to put a rack with mixed and non-arranged tubes in front of a robot. The robot autonomously performs recognition, reasoning, planning, manipulation, etc., and returns a rack with separated and arranged tubes. The system is simple-to-use, and there are no requests for expert knowledge in robotics. We expect such a system to play an important role in helping managing public health and hope similar systems could be extended to other clinical manipulation like handling mixers and pipettes in the future.

Network security has become an area of significant importance more than ever as highlighted by the eye-opening numbers of data breaches, attacks on critical infrastructure, and malware/ransomware/cryptojacker attacks that are reported almost every day. Increasingly, we are relying on networked infrastructure and with the advent of IoT, billions of devices will be connected to the internet, providing attackers with more opportunities to exploit. Traditional machine learning methods have been frequently used in the context of network security. However, such methods are more based on statistical features extracted from sources such as binaries, emails, and packet flows.

On the other hand, recent years witnessed a phenomenal growth in computer vision mainly driven by the advances in the area of convolutional neural networks. At a glance, it is not trivial to see how computer vision methods are related to network security. Nonetheless, there is a significant amount of work that highlighted how methods from computer vision can be applied in network security for detecting attacks or building security solutions. In this paper, we provide a comprehensive survey of such work under three topics; i) phishing attempt detection, ii) malware detection, and iii) traffic anomaly detection. Next, we review a set of such commercial products for which public information is available and explore how computer vision methods are effectively used in those products. Finally, we discuss existing research gaps and future research directions, especially focusing on how network security research community and the industry can leverage the exponential growth of computer vision methods to build much secure networked systems.

Reinforcement learning methods have been used to compute dialog policies from language-based interaction experiences. Efficiency is of particular importance in dialog policy learning, because of the considerable cost of interacting with people, and the very poor user experience from low-quality conversations. Aiming at improving the efficiency of dialog policy learning, we develop algorithm LHUA (Learning with Hindsight, User modeling, and Adaptation) that, for the first time, enables dialog agents to adaptively learn with hindsight from both simulated and real users. Simulation and hindsight provide the dialog agent with more experience and more (positive) reinforcements respectively. Experimental results suggest that, in success rate and policy quality, LHUA outperforms competitive baselines from the literature, including its no-simulation, no-adaptation, and no-hindsight counterparts.

Automated person re-identification in a multi-camera surveillance setup is very important for effective tracking and monitoring crowd movement. In the recent years, few deep learning based re-identification approaches have been developed which are quite accurate but time-intensive, and hence not very suitable for practical purposes. In this paper, we propose an efficient hierarchical re-identification approach in which color histogram based comparison is first employed to find the closest matches in the gallery set, and next deep feature based comparison is carried out using Siamese network. Reduction in search space after the first level of matching helps in achieving a fast response time as well as improving the accuracy of prediction by the Siamese network by eliminating vastly dissimilar elements. A silhouette part-based feature extraction scheme is adopted in each level of hierarchy to preserve the relative locations of the different body structures and make the appearance descriptors more discriminating in nature. The proposed approach has been evaluated on five public data sets and also a new data set captured by our team in our laboratory. Results reveal that it outperforms most state-of-the-art approaches in terms of overall accuracy.

In this paper, some useful necessary and sufficient conditions for the unique solution of the generalized absolute value equation (GAVE) $Ax-B|x|=b$ with $A, Bin mathbb{R}^{n imes n}$ from the optimization field are first presented, which cover the fundamental theorem for the unique solution of the linear system $Ax=b$ with $Ain mathbb{R}^{n imes n}$. Not only that, some new sufficient conditions for the unique solution of the GAVE are obtained, which are weaker than the previous published works.

Multi-view data capture permits free-viewpoint video (FVV) content creation. To this end, several users must capture video streams, calibrated in both time and pose, framing the same object/scene, from different viewpoints. New-generation network architectures (e.g. 5G) promise lower latency and larger bandwidth connections supported by powerful edge computing, properties that seem ideal for reliable FVV capture. We have explored this possibility, aiming to remove the need for bespoke synchronisation hardware when capturing a scene from multiple viewpoints, making it possible through off-the-shelf mobiles. We propose a novel and scalable data capture architecture that exploits edge resources to synchronise and harvest frame captures. We have designed an edge computing unit that supervises the relaying of timing triggers to and from multiple mobiles, in addition to synchronising frame harvesting. We empirically show the benefits of our edge computing unit by analysing latencies and show the quality of 3D reconstruction outputs against an alternative and popular centralised solution based on Unity3D.

Learning-based techniques are increasingly effective at controlling complex systems using data-driven models. However, most work done so far has focused on learning individual tasks or control laws. Hence, it is still a largely unaddressed research question how multiple tasks can be learned efficiently and simultaneously on the same system. In particular, no efficient state space exploration schemes have been designed for multi-task control settings. Using this research gap as our main motivation, we present an algorithm that approximates the smallest data set that needs to be collected in order to achieve high control performance for multiple learning-based control laws. We describe system uncertainty using a probabilistic Gaussian process model, which allows us to quantify the impact of potentially collected data on each learning-based controller. We then determine the optimal measurement locations by solving a stochastic optimization problem approximately. We show that, under reasonable assumptions, the approximate solution converges towards that of the exact problem. Additionally, we provide a numerical illustration of the proposed algorithm.

An important goal in reinforcement learning is to create agents that can quickly adapt to new goals while avoiding situations that might cause damage to themselves or their environments. One way agents learn is through exploration mechanisms, which are needed to discover new policies. However, in deep reinforcement learning, exploration is normally done by injecting noise in the action space. While performing well in many domains, this setup has the inherent risk that the noisy actions performed by the agent lead to unsafe states in the environment. Here we introduce a novel approach called Meta-Learned Instinctual Networks (MLIN) that allows agents to safely learn during their lifetime while avoiding potentially hazardous states. At the core of the approach is a plastic network trained through reinforcement learning and an evolved "instinctual" network, which does not change during the agent's lifetime but can modulate the noisy output of the plastic network. We test our idea on a simple 2D navigation task with no-go zones, in which the agent has to learn to approach new targets during deployment. MLIN outperforms standard meta-trained networks and allows agents to learn to navigate to new targets without colliding with any of the no-go zones. These results suggest that meta-learning augmented with an instinctual network is a promising new approach for safe AI, which may enable progress in this area on a variety of different domains.

Water quality has a direct impact on industry, agriculture, and public health. Algae species are common indicators of water quality. It is because algal communities are sensitive to changes in their habitats, giving valuable knowledge on variations in water quality. However, water quality analysis requires professional inspection of algal detection and classification under microscopes, which is very time-consuming and tedious. In this paper, we propose a novel multi-target deep learning framework for algal detection and classification. Extensive experiments were carried out on a large-scale colored microscopic algal dataset. Experimental results demonstrate that the proposed method leads to the promising performance on algal detection, class identification and genus identification.

Bayesian predictive coding is a putative neuromorphic method for acquiring higher-level neural representations to account for sensory input. Although originating in the neuroscience community, there are also efforts in the machine learning community to study these models. This paper reviews some of the more well known models. Our review analyzes module connectivity and patterns of information transfer, seeking to find general principles used across the models. We also survey some recent attempts to cast these models within a deep learning framework. A defining feature of Bayesian predictive coding is that it uses top-down, reconstructive mechanisms to predict incoming sensory inputs or their lower-level representations. Discrepancies between the predicted and the actual inputs, known as prediction errors, then give rise to future learning that refines and improves the predictive accuracy of learned higher-level representations. Predictive coding models intended to describe computations in the neocortex emerged prior to the development of deep learning and used a communication structure between modules that we name the Rao-Ballard protocol. This protocol was derived from a Bayesian generative model with some rather strong statistical assumptions. The RB protocol provides a rubric to assess the fidelity of deep learning models that claim to implement predictive coding.

Recently, the outbreak of Coronavirus Disease 2019 (COVID-19) has spread rapidly across the world. Due to the large number of affected patients and heavy labor for doctors, computer-aided diagnosis with machine learning algorithm is urgently needed, and could largely reduce the efforts of clinicians and accelerate the diagnosis process. Chest computed tomography (CT) has been recognized as an informative tool for diagnosis of the disease. In this study, we propose to conduct the diagnosis of COVID-19 with a series of features extracted from CT images. To fully explore multiple features describing CT images from different views, a unified latent representation is learned which can completely encode information from different aspects of features and is endowed with promising class structure for separability. Specifically, the completeness is guaranteed with a group of backward neural networks (each for one type of features), while by using class labels the representation is enforced to be compact within COVID-19/community-acquired pneumonia (CAP) and also a large margin is guaranteed between different types of pneumonia. In this way, our model can well avoid overfitting compared to the case of directly projecting highdimensional features into classes. Extensive experimental results show that the proposed method outperforms all comparison methods, and rather stable performances are observed when varying the numbers of training data.

Segmentation is a prerequisite yet challenging task for medical image analysis. In this paper, we introduce a novel deeply supervised active learning approach for finger bones segmentation. The proposed architecture is fine-tuned in an iterative and incremental learning manner. In each step, the deep supervision mechanism guides the learning process of hidden layers and selects samples to be labeled. Extensive experiments demonstrated that our method achieves competitive segmentation results using less labeled samples as compared with full annotation.

Artificial intelligence technology has made its way absolutely necessary in a variety of industries including the fitness industry. Human pose estimation is one of the important researches in the field of Computer Vision for the last few years. In this project, pose estimation and deep machine learning techniques are combined to analyze the performance and report feedback on the repetitions of performed exercises in real-time. Involving machine learning technology in the fitness industry could help the judges to count repetitions of any exercise during Weightlifting or CrossFit competitions.

By using dissipativity approach, we establish the stability condition for the feedback connection of a deterministic dynamical system $Sigma$ and a stochastic memoryless map $Psi$. After that, we extend the result to the class of large scale systems in which: $Sigma$ consists of many sub-systems; and $Psi$ consists of many "stochastic actuators" and "probability controllers" that control the actuator's output events. We will demonstrate the proposed approach by showing the design procedures to globally stabilize the manufacturing systems while locally balance the stock levels in any production process.

We analyze the computational complexity of motion planning through local "input/output" gadgets with separate entrances and exits, and a subset of allowed traversals from entrances to exits, each of which changes the state of the gadget and thereby the allowed traversals. We study such gadgets in the 0-, 1-, and 2-player settings, in particular extending past motion-planning-through-gadgets work to 0-player games for the first time, by considering "branchless" connections between gadgets that route every gadget's exit to a unique gadget's entrance. Our complexity results include containment in L, NL, P, NP, and PSPACE; as well as hardness for NL, P, NP, and PSPACE. We apply these results to show PSPACE-completeness for certain mechanics in Factorio, [the Sequence], and a restricted version of Trainyard, improving prior results. This work strengthens prior results on switching graphs and reachability switching games.

Convolutional neural networks (CNN) have shown promising results for end-to-end speech recognition, albeit still behind other state-of-the-art methods in performance. In this paper, we study how to bridge this gap and go beyond with a novel CNN-RNN-transducer architecture, which we call ContextNet. ContextNet features a fully convolutional encoder that incorporates global context information into convolution layers by adding squeeze-and-excitation modules. In addition, we propose a simple scaling method that scales the widths of ContextNet that achieves good trade-off between computation and accuracy. We demonstrate that on the widely used LibriSpeech benchmark, ContextNet achieves a word error rate (WER) of 2.1\%/4.6\% without external language model (LM), 1.9\%/4.1\% with LM and 2.9\%/7.0\% with only 10M parameters on the clean/noisy LibriSpeech test sets. This compares to the previous best published system of 2.0\%/4.6\% with LM and 3.9\%/11.3\% with 20M parameters. The superiority of the proposed ContextNet model is also verified on a much larger internal dataset.

Most optimization-based community detection approaches formulate the problem in a single or bi-objective framework. In this paper, we propose two variants of a three-objective formulation using a customized non-dominated sorting genetic algorithm III (NSGA-III) to find community structures in a network. In the first variant, named NSGA-III-KRM, we considered Kernel k means, Ratio cut, and Modularity, as the three objectives, whereas the second variant, named NSGA-III-CCM, considers Community score, Community fitness and Modularity, as three objective functions. Experiments are conducted on four benchmark network datasets. Comparison with state-of-the-art approaches along with decomposition-based multi-objective evolutionary algorithm variants (MOEA/D-KRM and MOEA/D-CCM) indicates that the proposed variants yield comparable or better results. This is particularly significant because the addition of the third objective does not worsen the results of the other two objectives. We also propose a simple method to rank the Pareto solutions so obtained by proposing a new measure, namely the ratio of the hyper-volume and inverted generational distance (IGD). The higher the ratio, the better is the Pareto set. This strategy is particularly useful in the absence of empirical attainment function in the multi-objective framework, where the number of objectives is more than two.

This paper investigates a class of optimal control problems associated with Markov processes with local state information. The decision-maker has only local access to a subset of a state vector information as often encountered in decentralized control problems in multi-agent systems. Under this information structure, part of the state vector cannot be observed. We leverage ab initio principles and find a new form of Bellman equations to characterize the optimal policies of the control problem under local information structures. The dynamic programming solutions feature a mixture of dynamics associated unobservable state components and the local state-feedback policy based on the observable local information. We further characterize the optimal local-state feedback policy using linear programming methods. To reduce the computational complexity of the optimal policy, we propose an approximate algorithm based on virtual beliefs to find a sub-optimal policy. We show the performance bounds on the sub-optimal solution and corroborate the results with numerical case studies.

In this work, we consider a group of robots working together to manipulate a rigid object to track a desired trajectory in $SE(3)$. The robots have no explicit communication network among them, and they do no know the mass or friction properties of the object, or where they are attached to the object. However, we assume they share data from a common IMU placed arbitrarily on the object. To solve this problem, we propose a decentralized adaptive control scheme wherein each agent maintains and adapts its own estimate of the object parameters in order to track a reference trajectory. We present an analysis of the controller's behavior, and show that all closed-loop signals remain bounded, and that the system trajectory will almost always (except for initial conditions on a set of measure zero) converge to the desired trajectory. We study the proposed controller's performance using numerical simulations of a manipulation task in 3D, and with hardware experiments which demonstrate our algorithm on a planar manipulation task. These studies, taken together, demonstrate the effectiveness of the proposed controller even in the presence of numerous unmodelled effects, such as discretization errors and complex frictional interactions.