Multi-Tenant Virtual GPUs for Optimising Performance of a Financial Risk Application

Graphics Processing Units (GPUs) are becoming popular accelerators in modern High-Performance Computing (HPC) clusters. Installing GPUs on each node of the cluster is not efficient resulting in high costs and power consumption as well as underutilisation of the accelerator. The research reported in this paper is motivated towards the use of few physical GPUs by providing cluster nodes access to remote GPUs on-demand for a financial risk application. We hypothesise that sharing GPUs between several nodes, referred to as multi-tenancy, reduces the execution time and energy consumed by an application. Two data transfer modes between the CPU and the GPUs, namely concurrent and sequential, are explored. The key result from the experiments is that multi-tenancy with few physical GPUs using sequential data transfers lowers the execution time and the energy consumed, thereby improving the overall performance of the application.

Robust indoor speaker recognition in a network of audio and video sensors

Situational awareness is achieved naturally by the human senses of sight and hearing in combination. Automatic scene understanding aims at replicating this human ability using microphones and cameras in cooperation. In this paper, audio and video signals are fused and integrated at different levels of semantic abstractions. We detect and track a speaker who is relatively unconstrained, i.e., free to move indoors within an area larger than the comparable reported work, which is usually limited to round table meetings. The system is relatively simple: consisting of just 4 microphone pairs and a single camera. Results show that the overall multimodal tracker is more reliable than single modality systems, tolerating large occlusions and cross-talk. System evaluation is performed on both single and multi-modality tracking. The performance improvement given by the audio–video integration and fusion is quantified in terms of tracking precision and accuracy as well as speaker diarisation error rate and precision–recall (recognition). Improvements vs. the closest works are evaluated: 56% sound source localisation computational cost over an audio only system, 8% speaker diarisation error rate over an audio only speaker recognition unit and 36% on the precision–recall metric over an audio–video dominant speaker recognition method.

Multi-Device Selection Scheduling in Non-Identically Distributed Fading Channels

Multiuser selection scheduling concept has been recently proposed in the literature in order to increase the multiuser diversity gain and overcome the significant feedback requirements for the opportunistic scheduling schemes. The main idea is that reducing the feedback overhead saves per-user power that could potentially be added for the data transmission. In this work, the authors propose to integrate the principle of multiuser selection and the proportional fair scheduling scheme. This is aimed especially at power-limited, multi-device systems in non-identically distributed fading channels. For the performance analysis, they derive closed-form expressions for the outage probabilities and the average system rate of the delay-sensitive and the delay-tolerant systems, respectively, and compare them with the full feedback multiuser diversity schemes. The discrete rate region is analytically presented, where the maximum average system rate can be obtained by properly choosing the number of partial devices. They optimise jointly the number of partial devices and the per-device power saving in order to maximise the average system rate under the power requirement. Through the authors’ results, they finally demonstrate that the proposed scheme leveraging the saved feedback power to add for the data transmission can outperform the full feedback multiuser diversity, in non-identical Rayleigh fading of devices’ channels.

Variation in pre-PCR processing of FFPE samples leads to discrepancies in BRAF and EGFR mutation detection: a diagnostic RING trial.

AIMS: Mutation detection accuracy has been described extensively; however, it is surprising that pre-PCR processing of formalin-fixed paraffin-embedded (FFPE) samples has not been systematically assessed in clinical context. We designed a RING trial to (i) investigate pre-PCR variability, (ii) correlate pre-PCR variation with EGFR/BRAF mutation testing accuracy and (iii) investigate causes for observed variation. METHODS: 13 molecular pathology laboratories were recruited. 104 blinded FFPE curls including engineered FFPE curls, cell-negative FFPE curls and control FFPE tissue samples were distributed to participants for pre-PCR processing and mutation detection. Follow-up analysis was performed to assess sample purity, DNA integrity and DNA quantitation. RESULTS: Rate of mutation detection failure was 11.9%. Of these failures, 80% were attributed to pre-PCR error. Significant differences in DNA yields across all samples were seen using analysis of variance (p

Achievable sum-rate of multiuser massive MIMO downlink in ricean fading channels

We investigate the achievable ergodic sum-rate of multi-user multiple-input multiple-output systems in Ricean fading channels. We first derive a lower bound on the average signal-to-leakage-and-noise ratio by utilizing the Mullen's inequality, which is then used to analyze the effect of channel mean information on the achievable sum-rate. With these results, a novel statistical-eigenmode space-division multipleaccess downlink transmission scheme is proposed. For this scheme, we derive an exact closed-form expression for the achievable ergodic sum-rate. Our results show that the achievable ergodic sum-rate converges to a saturation value in the high signal-to-noise ratio (SNR) region and reaches to a lower limit value in the lower Ricean K-factor range. In addition, we present tractable upper and lower bounds, which are shown to be tight for any SNR and Ricean K-factor value. Finally, the theoretical analysis is validated via numerical simulations.

Massive MIMO in sparse channels

Massive multi-user multiple-input multiple-output (MU-MIMO) systems are cellular networks where the base stations (BSs) are equipped with hundreds of antennas, N, and communicate with tens of mobile stations (MSs), K, such that, N ≫ K ≫ 1. Contrary to most prior works, in this paper, we consider the uplink of a single-cell massive MIMO system operating in sparse channels with limited scattering. This case is of particular importance in most propagation scenarios, where the prevalent Rayleigh fading assumption becomes idealistic. We derive analytical approximations for the achievable rates of maximum-ratio combining (MRC) and zero-forcing (ZF) receivers. Furthermore, we study the asymptotic behavior of the achievable rates for both MRC and ZF receivers, when N and K go to infinity under the condition that N/K → c ≥ 1. Our results indicate that the achievable rate of MRC receivers reaches an asymptotic saturation limit, whereas the achievable rate of ZF receivers grows logarithmically with the number of MSs.

Multipair massive MIMO full-duplex relaying with MRC/MRT processing

We consider a multipair relay channel, where multiple sources communicate with multiple destinations with the help of a full-duplex (FD) relay station (RS). All sources and destinations have a single antenna, while the RS is equipped with massive arrays. We assume that the RS estimates the channels by using training sequences transmitted from sources and destinations. Then, it uses maximum-ratio combining/maximum-ratio transmission (MRC/MRT) to process the signals. To significantly reduce the loop interference (LI) effect, we propose two massive MIMO processing techniques: i) using a massive receive antenna array; or ii) using a massive transmit antenna array together with very low transmit power at the RS. We derive an exact achievable rate in closed-form and evaluate the system spectral efficiency. We show that, by doubling the number of antennas at the RS, the transmit power of each source and of the RS can be reduced by 1.5 dB if the pilot power is equal to the signal power and by 3 dB if the pilot power is kept fixed, while maintaining a given quality-of-service. Furthermore, we compare FD and half-duplex (HD) modes and show that FD improves significantly the performance when the LI level is low.

Physical Layer Security with Threshold-Based Multiuser Scheduling in Multi-antenna Wireless Networks

In this paper, we consider a multiuser downlink wiretap network consisting of one base station (BS) equipped with AA antennas, NB single-antenna legitimate users, and NE single-antenna eavesdroppers over Nakagami-m fading channels. In particular, we introduce a joint secure transmission scheme that adopts transmit antenna selection (TAS) at the BS and explores threshold-based selection diversity (tSD) scheduling over legitimate users to achieve a good secrecy performance while maintaining low implementation complexity. More specifically, in an effort to quantify the secrecy performance of the considered system, two practical scenarios are investigated, i.e., Scenario I: the eavesdropper’s channel state information (CSI) is unavailable at the BS, and Scenario II: the eavesdropper’s CSI is available at the BS. For Scenario I, novel exact closed-form expressions of the secrecy outage probability are derived, which are valid for general networks with an arbitrary number of legitimate users, antenna configurations, number of eavesdroppers, and the switched threshold. For Scenario II, we take into account the ergodic secrecy rate as the principle performance metric, and derive novel closed-form expressions of the exact ergodic secrecy rate. Additionally, we also provide simple and asymptotic expressions for secrecy outage probability and ergodic secrecy rate under two distinct cases, i.e., Case I: the legitimate user is located close to the BS, and Case II: both the legitimate user and eavesdropper are located close to the BS. Our important findings reveal that the secrecy diversity order is AAmA and the slope of secrecy rate is one under Case I, while the secrecy diversity order and the slope of secrecy rate collapse to zero under Case II, where the secrecy performance floor occurs. Finally, when the switched threshold is carefully selected, the considered scheduling scheme outperforms other well known existing schemes in terms of the secrecy performance and complexity tradeoff

FPGA-based soft-core processors for image processing applications

With security and surveillance, there is an increasing need to process image data efficiently and effectively either at source or in a large data network. Whilst a Field-Programmable Gate Array (FPGA) has been seen as a key technology for enabling this, the design process has been viewed as problematic in terms of the time and effort needed for implementation and verification. The work here proposes a different approach of using optimized FPGA-based soft-core processors which allows the user to exploit the task and data level parallelism to achieve the quality of dedicated FPGA implementations whilst reducing design time. The paper also reports some preliminary
progress on the design flow to program the structure. An implementation for a Histogram of Gradients algorithm is also reported which shows that a performance of 328 fps can be achieved with this design approach, whilst avoiding the long design time, verification and debugging steps associated with conventional FPGA implementations.