Avian Influenza breakouts cause millions of dollars in damage each year globally, especially in Asian countries such as China and South Korea. The impact magnitude of a breakout directly correlates to time required to fully understand the influenza virus, particularly the interspecies pathogenicity. The procedure requires laboratory tests that require resources typically lacking in a breakout emergency. In this study, we propose new quantitative methods utilizing machine learning and deep learning to correctly classify host species given raw DNA sequence data of the influenza virus, and provide probabilities for each classification. The best deep learning models achieve top-1 classification accuracy of 47%, and top-3 classification accuracy of 82%, on a dataset of 11 host species classes.
When simultaneous wireless information and power transfer is carried out, a fundamental tradeoff between achievable rate and harvested energy exists because the received power is used for two different purposes. The tradeoff is well characterized by the rate-energy region, and several techniques have been proposed to improve the achievable rate-energy region. However, the existing techniques still have a considerable loss in either energy or rate and thus the known achievable rate-energy regions are far from the ideal one. Deriving tight upper and lower bounds on the rate-energy region of our proposed scheme, we prove that the rate-energy region can be expanded almost to the ideal upper bound. Contrary to the existing techniques, in the proposed scheme, the information decoding circuit not only extracts amplitude and phase information but also combines the extracted information with the amplitude information obtained from the rectified signal. Consequently, the required energy for decoding can be minimized, and thus the proposed scheme achieves a near-optimal rate-energy region, which implies that the fundamental tradeoff in the achievable rate-energy region is nearly eliminated. To practically account for the theoretically achievable rate-energy region, we also present practical examples with an $M$-ary multi-level circular QAM with Gaussian maximum likelihood detection.
Dec 07 2017 cs.DC
Convolutional Neural Networks (CNNs) have shown a great deal of success in diverse application domains including computer vision, speech recognition, and natural language processing. However, as the size of datasets and the depth of neural network architectures continue to grow, it is imperative to design high-performance and energy-efficient computing hardware for training CNNs. In this paper, we consider the problem of designing specialized CPU-GPU based heterogeneous manycore systems for energy-efficient training of CNNs. It has already been shown that the typical on-chip communication infrastructures employed in conventional CPU-GPU based heterogeneous manycore platforms are unable to handle both CPU and GPU communication requirements efficiently. To address this issue, we first analyze the on-chip traffic patterns that arise from the computational processes associated with training two deep CNN architectures, namely, LeNet and CDBNet, to perform image classification. By leveraging this knowledge, we design a hybrid Network-on-Chip (NoC) architecture, which consists of both wireline and wireless links, to improve the performance of CPU-GPU based heterogeneous manycore platforms running the above-mentioned CNN training workloads. The proposed NoC achieves 1.8x reduction in network latency and improves the network throughput by a factor of 2.2 for training CNNs, when compared to a highly-optimized wireline mesh NoC. For the considered CNN workloads, these network-level improvements translate into 25% savings in full-system energy-delay-product (EDP). This demonstrates that the proposed hybrid NoC for heterogeneous manycore architectures is capable of significantly accelerating training of CNNs while remaining energy-efficient.
Block traces are widely used for system studies, model verifications, and design analyses in both industry and academia. While such traces include detailed block access patterns, existing trace-driven research unfortunately often fails to find true-north due to a lack of runtime contexts such as user idle periods and system delays, which are fundamentally linked to the characteristics of target storage hardware. In this work, we propose TraceTracker, a novel hardware/software co-evaluation method that allows users to reuse a broad range of the existing block traces by keeping most their execution contexts and user scenarios while adjusting them with new system information. Specifically, our TraceTracker's software evaluation model can infer CPU burst times and user idle periods from old storage traces, whereas its hardware evaluation method remasters the storage traces by interoperating the inferred time information, and updates all inter-arrival times by making them aware of the target storage system. We apply the proposed co-evaluation model to 577 traces, which were collected by servers from different institutions and locations a decade ago, and revive the traces on a high-performance flash-based storage array. The evaluation results reveal that the accuracy of the execution contexts reconstructed by TraceTracker is on average 99% and 96% with regard to the frequency of idle operations and the total idle periods, respectively.
Jun 27 2017 cs.CV
Data association problems are an important component of many computer vision applications, with multi-object tracking being one of the most prominent examples. A typical approach to data association involves finding a graph matching or network flow that minimizes a sum of pairwise association costs, which are often either hand-crafted or learned as linear functions of fixed features. In this work, we demonstrate that it is possible to learn features for network-flow-based data association via backpropagation, by expressing the optimum of a smoothed network flow problem as a differentiable function of the pairwise association costs. We apply this approach to multi-object tracking with a network flow formulation. Our experiments demonstrate that we are able to successfully learn all cost functions for the association problem in an end-to-end fashion, which outperform hand-crafted costs in all settings. The integration and combination of various sources of inputs becomes easy and the cost functions can be learned entirely from data, alleviating tedious hand-designing of costs.
Apr 19 2017 cs.AR
Storage-class memory (SCM) combines the benefits of a solid-state memory, such as high-performance and robustness, with the archival capabilities and low cost of conventional hard-disk magnetic storage. Among candidate solid-state nonvolatile memory technologies that could potentially be used to construct SCM, flash memory is a well-established technology and have been widely used in commercially available SCM incarnations. Flash-based SCM enables much better tradeoffs between performance, space and power than disk-based systems. However, write endurance is a significant challenge for a flash-based SCM (each act of writing a bit may slightly damage a cell, so one flash cell can be written 10^4--10^5 times, depending on the flash technology, before it becomes unusable). This is a well-documented problem and has received a lot of attention by manufactures that are using some combination of write reduction and wear-leveling techniques for achieving longer lifetime. In an effort to improve flash lifetime, first, by quantifying data longevity in an SCM, we show that a majority of the data stored in a solid-state SCM do not require long retention times provided by flash memory (i.e., up to 10 years in modern devices); second, by exploiting retention time relaxation, we propose a novel mechanism, called Dense-SLC (D-SLC), which enables us perform multiple writes into a cell during each erase cycle for lifetime extension; and finally, we discuss the required changes in the flash management software (FTL) in order to use this characteristic for extending the lifetime of the solid-state part of an SCM. Using an extensive simulation-based analysis of a flash-based SCM, we demonstrate that D-SLC is able to significantly improve device lifetime (between 5.1X and 8.6X) with no performance overhead and also very small changes at the FTL software.
Apr 17 2017 cs.CV
We introduce a Deep Stochastic IOC RNN Encoderdecoder framework, DESIRE, for the task of future predictions of multiple interacting agents in dynamic scenes. DESIRE effectively predicts future locations of objects in multiple scenes by 1) accounting for the multi-modal nature of the future prediction (i.e., given the same context, future may vary), 2) foreseeing the potential future outcomes and make a strategic prediction based on that, and 3) reasoning not only from the past motion history, but also from the scene context as well as the interactions among the agents. DESIRE achieves these in a single end-to-end trainable neural network model, while being computationally efficient. The model first obtains a diverse set of hypothetical future prediction samples employing a conditional variational autoencoder, which are ranked and refined by the following RNN scoring-regression module. Samples are scored by accounting for accumulated future rewards, which enables better long-term strategic decisions similar to IOC frameworks. An RNN scene context fusion module jointly captures past motion histories, the semantic scene context and interactions among multiple agents. A feedback mechanism iterates over the ranking and refinement to further boost the prediction accuracy. We evaluate our model on two publicly available datasets: KITTI and Stanford Drone Dataset. Our experiments show that the proposed model significantly improves the prediction accuracy compared to other baseline methods.
Jul 05 2016 cs.CR
In the last several decades, the automotive industry has come to incorporate the latest Information and Communications (ICT) technology, increasingly replacing mechanical components of vehicles with electronic components. These electronic control units (ECUs) communicate with each other in an in-vehicle network that makes the vehicle both safer and easier to drive. Controller Area Networks (CANs) are the current standard for such high quality in-vehicle communication. Unfortunately, however, CANs do not currently offer protection against security attacks. In particular, they do not allow for message authentication and hence are open to attacks that replay ECU messages for malicious purposes. Applying the classic cryptographic method of message authentication code (MAC) is not feasible since the CAN data frame is not long enough to include a sufficiently long MAC to provide effective authentication. In this paper, we propose a novel identification method, which works in the physical layer of an in-vehicle CAN network. Our method identifies ECUs using inimitable characteristics of signals enabling detection of a compromised or alien ECU being used in a replay attack. Unlike previous attempts to address security issues in the in-vehicle CAN network, our method works by simply adding a monitoring unit to the existing network, making it deployable in current systems and compliant with required CAN standards. Our experimental results show that the bit string and classification algorithm that we utilized yielded more accurate identification of compromised ECUs than any other method proposed to date. The false positive rate is more than 2 times lower than the method proposed by P.-S. Murvay et al. This paper is also the first to identify potential attack models that systems should be able to detect.
We solve a sum rate maximization problem of full-duplex (FD) multiuser multiple-input multiple-output (MU-MIMO) systems. Since additional self-interference (SI) in the uplink channel and co-channel interference (CCI) in the downlink channel are coupled in FD communication, the downlink and uplink multiuser beamforming vectors are required to be jointly designed. However, the joint optimization problem is non-convex and hard to solve due to the coupled effect. To properly address the coupled design issue, we reformulate the problem into an equivalent uplink channel problem, using the uplink and downlink channel duality known as MAC-BC duality. Then, using minorization maximization (MM) algorithm based on an affine approximation, we obtain a solution for the reformulated problem. In addition, without any approximation and thus performance degradation, we develop an alternating algorithm based on iterative water-filling (IWF) to solve the non-convex problem. The proposed algorithms warrant fast convergence and low computational complexity.
We consider the full-duplex (FD) two-way amplify-and-forward relay system with imperfect cancelation of loopback self-interference (SI) and investigate joint design of relay and receive beamforming for minimizing the mean square error under a relay transmit power constraint. Due to loopback channel estimation error and limitation of analog-to-digital converter, the loopback SI cannot be completely canceled. Multiple antennas at the relay can help loopback SI suppression but beamforming is required to balance between the residual SI suppression and the desired signal transmission. Moreover, the relay beamforming matrix should be updated every time slot because the residual SI in the previous time slot is amplified by the current beamforming matrix and added to the received signals from the two sources in the current time slot. We derive the optimally balanced relay beamforming and receive beamforming matrices in closed form based on minimum mean square error, taking into account the propagation of the residual loopback SI from the first to the current time slot. We also propose beamforming design using only the channels of the m latest time slots, not from the first time slot. Based on our numerical results, we also identify when FD is beneficial and propose selection between FD and half-duplex according to signal-to-noise ratio and interference-to-noise ratio.
In wireless cloud storage systems, the recovery failure probability depends on not only wireless channel conditions but also storage size of each distributed storage node. For an efficient utilization of limited storage capacity and the performance characterization of allocation strategies, we asymptotically analyze the recovery failure probability of a wireless cloud storage system with a sum storage capacity constraint for both high SNR regime and low SNR regime. Then, we find the optimal storage allocation strategy across distributed storage nodes in terms of the asymptotic recovery failure probability. Our analysis reveals that the maximal symmetric allocation is optimal for high SNR regime and the minimal allocation (with $\lfloor T\rfloor$ complete storage nodes and an incomplete storage node) is optimal for low SNR regime, where $T$ is the sum storage capacity. Based on the numerical investigation, we also show that in intermediate SNR regime, a balance allocation between the minimal allocation and the maximal symmetric allocation would not be required if we select one between them according to SNR.
Content delivery success in wireless caching helper networks depends mainly on cache-based channel selection diversity and network interference. For given channel fading and network geometry, both channel selection diversity and network interference dynamically vary according to what and how the caching helpers cache at their finite storage space. We study probabilistic content placement (or caching placement) to desirably control cache-based channel selection diversity and network interference in a stochastic wireless caching helper network, with sophisticated considerations of wireless fading channels, interactions among multiple users such as interference and loads at caching helpers, and arbitrary memory size. Using stochastic geometry, we derive optimal caching probabilities in closed form to maximize the average success probability of content delivery and propose an efficient algorithm to find the solution in a noise-limited network. In an interference-limited network, based on a lower bound of the average success probability of content delivery, we find near-optimal caching probabilities in closed form to control the channel selection diversity and the network interference. We numerically verify that the proposed content placement is superior to other comparable content placement strategies.
This paper investigates optimal caching placement for wireless femto-caching network. The average bit error rate (BER) is formulated as a function of caching placement under wireless fading. To minimize the average BER, we propose a greedy algorithm finding optimal caching placement with low computational complexity. Exploiting the property of the optimal caching placement which we derive, the proposed algorithm can be performed over considerably reduced search space. Contrary to the optimal caching placement without consideration of wireless fading aspects, we reveal that optimal caching placement can be reached by balancing a tradeoff between two different gains: file diversity gain and channel diversity gain. Moreover, we also identify the conditions that the optimal placement can be found without running the proposed greedy algorithm and derive the corresponding optimal caching placement in closed form.
Apr 19 2016 cs.CV
In CNN-based object detection methods, region proposal becomes a bottleneck when objects exhibit significant scale variation, occlusion or truncation. In addition, these methods mainly focus on 2D object detection and cannot estimate detailed properties of objects. In this paper, we propose subcategory-aware CNNs for object detection. We introduce a novel region proposal network that uses subcategory information to guide the proposal generating process, and a new detection network for joint detection and subcategory classification. By using subcategories related to object pose, we achieve state-of-the-art performance on both detection and pose estimation on commonly used benchmarks.
This paper studies the interference channel with a cognitive relay (ICCR) under delayed feedback. Three types of delayed feedback are studied: delayed channel state information at the transmitter (CSIT), delayed output feedback, and delayed Shannon feedback. Outer bounds are derived for the DoF region of the two-user multiple-input multiple-output (MIMO) ICCR with delayed feedback as well as without feedback. For the single-input single-output (SISO) scenario, optimal schemes are proposed based on retrospective interference alignment. It is shown that while a cognitive relay without feedback cannot extend the sum-DoF beyond $1$ in the two-user SISO interference channel, delayed feedback in the same scenario can extend the sum-DoF to $4/3$. For the MIMO case, achievable schemes are obtained via extensions of retrospective interference alignment, leading to DoF regions that meet the respective upper bounds.
Apr 10 2015 cs.CV
In this paper, we focus on the two key aspects of multiple target tracking problem: 1) designing an accurate affinity measure to associate detections and 2) implementing an efficient and accurate (near) online multiple target tracking algorithm. As the first contribution, we introduce a novel Aggregated Local Flow Descriptor (ALFD) that encodes the relative motion pattern between a pair of temporally distant detections using long term interest point trajectories (IPTs). Leveraging on the IPTs, the ALFD provides a robust affinity measure for estimating the likelihood of matching detections regardless of the application scenarios. As another contribution, we present a Near-Online Multi-target Tracking (NOMT) algorithm. The tracking problem is formulated as a data-association between targets and detections in a temporal window, that is performed repeatedly at every frame. While being efficient, NOMT achieves robustness via integrating multiple cues including ALFD metric, target dynamics, appearance similarity, and long term trajectory regularization into the model. Our ablative analysis verifies the superiority of the ALFD metric over the other conventional affinity metrics. We run a comprehensive experimental evaluation on two challenging tracking datasets, KITTI and MOT datasets. The NOMT method combined with ALFD metric achieves the best accuracy in both datasets with significant margins (about 10% higher MOTA) over the state-of-the-arts.
The input numerical aperture (NA) of multimode fiber (MMF) can be effectively increased by placing turbid media at the input end of the MMF. This provides the potential for high-resolution imaging through the MMF. While the input NA is increased, the number of propagation modes in the MMF and hence the output NA remains the same. This makes the image reconstruction process underdetermined and may limit the quality of the image reconstruction. In this paper, we aim to improve the signal to noise ratio (SNR) of the image reconstruction in imaging through MMF. We notice that turbid media placed in the input of the MMF transforms the incoming waves into a better format for information transmission and information extraction. We call this transformation as holistic random (HR) encoding of turbid media. By exploiting the HR encoding, we make a considerable improvement on the SNR of the image reconstruction. For efficient utilization of the HR encoding, we employ sparse representation (SR), a relatively new signal reconstruction framework when it is provided with a HR encoded signal. This study shows for the first time to our knowledge the benefit of utilizing the HR encoding of turbid media for recovery in the optically underdetermined systems where the output NA of it is smaller than the input NA for imaging through MMF.
What will 5G be? What it will not be is an incremental advance on 4G. The previous four generations of cellular technology have each been a major paradigm shift that has broken backwards compatibility. And indeed, 5G will need to be a paradigm shift that includes very high carrier frequencies with massive bandwidths, extreme base station and device densities and unprecedented numbers of antennas. But unlike the previous four generations, it will also be highly integrative: tying any new 5G air interface and spectrum together with LTE and WiFi to provide universal high-rate coverage and a seamless user experience. To support this, the core network will also have to reach unprecedented levels of flexibility and intelligence, spectrum regulation will need to be rethought and improved, and energy and cost efficiencies will become even more critical considerations. This paper discusses all of these topics, identifying key challenges for future research and preliminary 5G standardization activities, while providing a comprehensive overview of the current literature, and in particular of the papers appearing in this special issue.
In this paper, we propose opportunistic jammer selection in a wireless security system for increasing the secure degrees of freedom (DoF) between a transmitter and a legitimate receiver (say, Alice and Bob). There is a jammer group consisting of $S$ jammers among which Bob selects $K$ jammers. The selected jammers transmit independent and identically distributed Gaussian signals to hinder the eavesdropper (Eve). Since the channels of Bob and Eve are independent, we can select the jammers whose jamming channels are aligned at Bob, but not at Eve. As a result, Eve cannot obtain any DoF unless it has more than $KN_j$ receive antennas, where $N_j$ is the number of jammer's transmit antenna each, and hence $KN_j$ can be regarded as defensible dimensions against Eve. For the jamming signal alignment at Bob, we propose two opportunistic jammer selection schemes and find the scaling law of the required number of jammers for target secure DoF by a geometrical interpretation of the received signals.
In this paper, we consider a multiple-input multiple-output broadcast channel with limited feedback where all users share the feedback rates. Firstly, we find the optimal feedback rate sharing strategy using zero-forcing transmission scheme at the transmitter and random vector quantization at each user. We mathematically prove that equal sharing of sum feedback size among all users is the optimal strategy in the low signal-to-noise ratio (SNR) region, while allocating whole feedback size to a single user is the optimal strategy in the high SNR region. For the mid-SNR region, we propose a simple numerical method to find the optimal feedback rate sharing strategy based on our analysis and show that the equal allocation of sum feedback rate to a partial number of users is the optimal strategy. It is also shown that the proposed simple numerical method can be applicable to finding the optimal feedback rate sharing strategy when different path losses of the users are taken into account. We show that our proposed feedback rate sharing scheme can be extended to the system with stream control and is still useful for the systems with other techniques such as regularized zero-forcing and spherical cap codebook.
We propose new ergodic interference alignment techniques for $K$-user interference channels with delayed feedback. Two delayed feedback scenarios are considered -- delayed channel information at transmitter (CIT) and delayed output feedback. It is proved that the proposed techniques achieve total $2K/(K+2)$ DoF which is higher than that by the retrospective interference alignment for the delayed feedback scenarios.
Mar 21 2012 cs.PL
Generic programming (GP) is an increasingly important trend in programming languages. Well-known GP mechanisms, such as type classes and the C++0x concepts proposal, usually combine two features: 1) a special type of interfaces; and 2) implicit instantiation of implementations of those interfaces. Scala implicits are a GP language mechanism, inspired by type classes, that break with the tradition of coupling implicit instantiation with a special type of interface. Instead, implicits provide only implicit instantiation, which is generalized to work for any types. This turns out to be quite powerful and useful to address many limitations that show up in other GP mechanisms. This paper synthesizes the key ideas of implicits formally in a minimal and general core calculus called the implicit calculus, and it shows how to build source languages supporting implicit instantiation on top of it. A novelty of the calculus is its support for partial resolution and higher-order rules (a feature that has been proposed before, but was never formalized or implemented). Ultimately, the implicit calculus provides a formal model of implicits, which can be used by language designers to study and inform implementations of similar mechanisms in their own languages.
This paper investigates how multiuser dimensions can effectively be exploited for target degrees of freedom (DoF) in interfering broadcast channels (IBC) consisting of K-transmitters and their user groups. First, each transmitter is assumed to have a single antenna and serve a singe user in its user group where each user has receive antennas less than K. In this case, a K-transmitter single-input multiple-output (SIMO) interference channel (IC) is constituted after user selection. Without help of multiuser diversity, K-1 interfering signals cannot be perfectly removed at each user since the number of receive antennas is smaller than or equal to the number of interferers. Only with proper user selection, non-zero DoF per transmitter is achievable as the number of users increases. Through geometric interpretation of interfering channels, we show that the multiuser dimensions have to be used first for reducing the DoF loss caused by the interfering signals, and then have to be used for increasing the DoF gain from its own signal. The sufficient number of users for the target DoF is derived. We also discuss how the optimal strategy of exploiting multiuser diversity can be realized by practical user selection schemes. Finally, the single transmit antenna case is extended to the multiple-input multiple-output (MIMO) IBC where each transmitter with multiple antennas serves multiple users.
In this paper, we propose opportunistic interference alignment (OIA) schemes for three-transmitter multiple-input multiple-output (MIMO) interference channels (ICs). In the proposed OIA, each transmitter has its own user group and selects a single user who has the most aligned interference signals. The user dimensions provided by multiple users are exploited to align interfering signals. Contrary to conventional IA, perfect channel state information of all channel links is not required at the transmitter, and each user just feeds back one scalar value to indicate how well the interfering channels are aligned. We prove that each transmitter can achieve the same degrees of freedom (DoF) as the interference free case via user selection in our system model that the number of receive antennas is twice of the number of transmit antennas. Using the geometric interpretation, we find the required user scaling to obtain an arbitrary non-zero DoF. Two OIA schemes are proposed and compared with various user selection schemes in terms of achievable rate/DoF and complexity.
This paper analyzes the multiuser diversity gain in a cognitive radio (CR) system where secondary transmitters opportunistically utilize the spectrum licensed to primary users only when it is not occupied by the primary users. To protect the primary users from the interference caused by the missed detection of primary transmissions in the secondary network, minimum average throughput of the primary network is guaranteed by transmit power control at the secondary transmitters. The traffic dynamics of a primary network are also considered in our analysis. We derive the average achievable capacity of the secondary network and analyze its asymptotic behaviors to characterize the multiuser diversity gains in the CR system.