A processor architecture for Horizon

Spatial and temporal data such as multiple data streams often have concept drift problems, which refers to changes of the data distributions over time. Once concept drift occurs, a stationary machine learning predictor will probably be invalid because the testing data have a different data distribution from that of the training data. Recent studies in data streams aim to address this issue by concept drift adaptation techniques. Concept drift adaptation methods update the predictor by time and have been validated to provide accurate real-time prediction for a single data stream. However, handling multiple relevant data streams is still an unsolved challenge in this area, considering many real-world applications generate multiple data streams that are simultaneously evolving. To fill this gap, we here present a predicting network for multiple data streams, named MuNet. MuNet leverages the dependency between data streams to efficiently lower down the computational cost of real-time predictions for multiple data streams. In MuNet, an online-learned Bayesian neural network (BNN) is designed as a connector between streams. The BNN-connector can continuously use the real-time information from only one base stream to correct the stationary predictor for the other streams, which avoids the high cost caused by repeated adaptation.

9 - Parallel Processors

Parallel processing with minicomputers

Three-dimensional image registration is computationally intensive and therefore slow. Parallelizing the primitive computation processes involved in registration can decrease the time. Our approach for improving image processing speed is to utilize a massively parallel single instruction stream -- multiple data stream (SIMD) computer. This paper presents the performance results comparing a massively parallel architecture to a four processor vector/parallel system and a conventional serial computer.

The parallel implementation aspects of the Viterbi algorithm, which gives the maximum likelihood solution to decoding conventional codes, are studied. The interconnection of the trellis diagram of a binary convolutional code is similar to the shuffle exchange. This similarity is exploited to present a parallel version of the Viterbi algorithm. The proposed algorithm avoids tasks such as appending and exchanging the survivor lists that are considered pragmatic issues to be accommodated in implementing the Viterbi algorithm. This is done by introducing an additional back-tracking procedure which is almost identical to the forward path where branch and path metrics are computed. The parallel Viterbi algorithm is implemented on the connection machine which is a SIMD (single instruction stream and multiple data stream) parallel machine. >

We consider a single-cell multiple-input multiple-output (MIMO) downlink channel where linear transmission and reception strategy is employed. The base station (BS) transmitter is equipped with a scheduler using a simple opportunistic beamforming strategy, which associates an intended user for each of the transmitted data streams. For the case when the channel of the scheduled users is available at the BS, we propose a general method for joint design of the transmit and the receive beamformers according to different optimization criteria, including weighted sum rate maximization, weighted sum mean square error minimization, minimum signal-to-interference-plus-noise ratio (SINR) maximization and sum power minimization under a minimum SINR constraint. The proposed method can handle multiple antennas at the BS and at the mobile user with single and/or multiple data streams per scheduled user. The optimization problems encountered in the beamformer design (e.g., covariance rank constraint) are not convex in general. Therefore, the problem of finding the global optimum is intrinsically nontractable. However, by exploiting the uplink-downlink SINR duality, we decompose the original optimization problem as a series of simpler optimization problems which can be efficiently solved by using standard convex optimization tools. Even though each subproblem is optimally solved, there is no guarantee that the global optimum has been found due to the nonconvexity of the problem. However, the simulations show that the algorithms converge fast to a solution, which can be a local optimum, but is still efficient.

This paper investigates simple user selection strategies in a multiuser downlink system with a single transmitter, multiple legitimate receivers, and a single eavesdropper, where all nodes are equipped with multiple antennas. No information regarding the eavesdropper is presumed at the transmitter, and we examine the MIMO downlink channel with either single or multiple data streams per receiver. The transmitter uses linear beamforming techniques based on generalized zero-forcing with a corresponding minimum SINR or rate requirement per receiver. In both cases the information signal is transmitted with just enough power to guarantee the desired Quality-of-Service (QoS) at the desired receivers, while the remainder of the power is used to broadcast artificial noise that selectively degrades the passive eavesdropper's signal. Numerical simulations displaying the power consumption and the number of selected users provided by each selection scheme demonstrate the effectiveness of the proposed schedulers in ensuring physical-layer security.

In the last decade, researchers have recognized the need of an increased attention to a type of knowledge discovery applications where the data analyzed is not finite, but streams into the system continuously and endlessly. Data streams are ubiquitous, entering almost every area of modern life. As a result, processing, managing and learning from multiple data streams have become important and challenging tasks for the data mining, database and machine learning communities. Although a substantial body of algorithms for processing and learning from data streams has been developed, most of the work is focused on one-dimensional numerical data streams (time series) or a single multi-dimensional data stream. Only few of the existing solutions consider the most realistic scenario where data can be incomplete, correlated with other streams of information and can arrive from multiple heterogeneous sources.This paper discusses the requirements and the difficulties for learning from multiple multi-dimensional data streams inter-linked according to a pre-defined semantic schema (multi-relational data streams). The main research problem is to develop a time-efficient, resource-aware methodology for linking and exploring the information which is arriving independently and in an asynchronous way from its respective sources. The resulting framework has to enable, at any time error-bounded approximate answers to aggregate queries commonly issued in the process of multi-relational data mining. In particular we focus on the task of learning regression trees and their variants (model trees, option trees, multi-target trees) from multiple correlated streaming sources. To the best of our knowledge, no other work has previously addressed the problem of learning regression trees from multi-relational data streams.

Many current and state-of-the-art deep learning models for accurate image segmentation are based on the U-Net architecture, a convolutional neural network designed for biomedical applications. Despite its widespread adoption in the medical imaging community, U-Net has two major limitations. First, due to its deep structure and the large number of filters used, the number of parameters and Floating-Point Operations Per Second (FLOPS) are high. This results in high computational complexity and demands a large memory size, making real-time implementation and deployment of U-Net models challenging. Second, the base U-Net model only uses a single type of kernel (i.e., 3×3) throughout the network for all convolution operations. Feature extraction using a single spatial extent throughout the network is suitable if the size, location, and shape of the salient regions remain static in all the images of a dataset, which is not necessarily the case in medical imaging. To address these two limitations, we propose an Efficient Multi-Encoder-Decoder based UNet (EMED-UNet), a novel architecture for efficient medical image segmentation. We evaluated our network on four medical imaging datasets: Montgomery County, Shenzhen CXR, COVID-19 CT LS, and BraTS (Brain Tumor Segmentation) dataset. EMED-UNet outperforms the U-Net and its variants in terms of accuracy, with around 77% reduction in parameters, a 60% reduction in FLOPS, and a 79.2% reduction in memory usage (all as compared to U-Net). The results demonstrate that EMED-UNet is a lightweight and accurate model for image segmentation that substantially improves upon the U-Net base model and is more feasible to deploy, given its decreased computational cost.

Multiple, high-level operators are assigned to a general operand. The operators are implemented with individual micro-processors and are "attached" dynamically to the memory location representing the "current" value of the operand. The operators then asynchronously use each new operand value as it is stored and perform their operations in parallel. The proposed architecture represents a true M I S D (Multiple Instruction Stream - Single Data Stream) computer. Its architecture can provide effective parallelism and reduced programming complexity for a large class of both numeric and non-numeric computer problems.

We are proposing a new data path design for a position emission tomography (PET) data acquisition system (DAQ). The proposed design can transmit, receive, and process all types of data through a single data stream. Instead of having separate path for data, clock and control signals individually, which is the general case of a conventional DAQ system, the proposed system uses a packet concept to identify and process different data types in a single fast data stream.

The Multiple Associative Computing (MASC) parallelmodel is a generalization model of an AssociativeComputing (ASC) parallel model designed to supportmultiple ASC data parallel threads by using controlparallelism.

FFT algorithms for SIMD parallel processing systems

As the complexity of speech recognition systems increases, conventional computers are unable to perform all the operations quickly enough to process the speech input in real time. This paper examines the use of a single instruction stream - multiple data stream (SIMD) parallel architecture to reduce the computation time for a dynamic time warping template matching based isolated word recognition system. Each of the components of the recognition system was written as an SMD parallel algorithm. The component SIMD algorithms and the complete word recognition system were simulated. The simulations showed that with 100 processing elements, each an 8 MHz MC68000, the speech system can perform isolated word recognition over a large vocabulary in real time. This real-time speech system used a 20 KHz sampling rate, 16 bits per sample, 100 samples per frame, 8 LPC coefficients per frame, 16 bits per LPC coefficient, and a 1,000 word vocabulary.

The application of optical remote sensing images (ORSIs) is prevalent in many fields. Accordingly, ORSI-oriented salient object detection (SOD) has attracted more attention in recent years. However, yet many previously proposed methods present appealing performance in natural scene images (NSIs), they are difficult to be directly extended to remote sensing images due to the more complex scenes, such as blended backgrounds and diversiform topological shapes. Most specifically designed models often fail to achieve satisfactory results due to the weak usage of edge information and the ignorance of attention loss. Besides, computational inefficiency often causes poor applicability. To solve these problems, we propose a new model, namely, Bidimensional Attention and Full-stage Semantic Guidance Network (BAFS-Net), containing an edge guidance branch and a mainstream detection branch. Concretely, edge guidance generates boundary information, in which supervision with border labels is imposed to highlight the salient regions and plays a complementary role on the main branch. The mainstream detection branch involves two important components, i.e., bidimensional attention modules (BAMs) and semantic-guided fusion modules (SGFMs). Between these two, BAM uniformly assembles channel and spatial attention in an efficient and rational manner, addressing the open issue of dimensionwisely attention computation. SGFM hammers at the fusion of high-level features and low-level features. Moreover, the semantic maps are employed to interact with SGFM in full stages. Our approach surpasses most state-of-the-art RSI-SOD methods proposed in recent years, with respect to the accuracy, parameter size, computational cost, and floating point operations per second (FLOPS). The code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/ZhengJianwei2/BAFS-Net</uri> .