Intel (Israel)

Geometric deep learning is an umbrella term for emerging techniques attempting to generalize (structured) deep neural models to non-Euclidean domains, such as graphs and manifolds. The purpose of this article is to overview different examples of geometric deep-learning problems and present available solutions, key difficulties, applications, and future research directions in this nascent field.

Designed to accelerate multimedia and communications software, MMX technology improves performance by introducing data types and instructions to the IA that exploit the parallelism in these applications. MMX technology extends the Intel architecture (IA) to improve the performance of multimedia, communications, and other numeric-intensive applications. It uses a SIMD (single-instruction, multiple-data) technique to exploit the parallelism inherent in many algorithms, producing full application performance of 1.5 to 2 times faster than the same applications run on the same processor without MMX. The extension also maintains full compatibility with existing IA microprocessors, operating systems, and applications while providing new instructions and data types that applications can use to achieve a higher level of performance on the host CPU.

IEEE 802.16m and 3GPP LTE-Advanced are the two evolving standards targeting 4G wireless systems. In both standards, multiple-input multiple-output antenna technologies play an essential role in meeting the 4G requirements. The application of MIMO technologies is one of the most crucial distinctions between 3G and 4G. It not only enhances the conventional point-to-point link, but also enables new types of links such as downlink multiuser MIMO. A large family of MIMO techniques has been developed for various links and with various amounts of available channel state information in both IEEE 802.16e/m and 3GPP LTE/LTE-Advanced. In this article we provide a survey of the MIMO techniques in the two standards. The MIMO features of the two are compared, and the engineering considerations are depicted.

Current-generation Deep Neural Networks (DNNs), such as AlexNet and VGG, rely heavily on dense floating-point matrix multiplication (GEMM), which maps well to GPUs (regular parallelism, high TFLOP/s). Because of this, GPUs are widely used for accelerating DNNs. Current FPGAs offer superior energy efficiency (Ops/Watt), but they do not offer the performance of today's GPUs on DNNs. In this paper, we look at upcoming FPGA technology advances, the rapid pace of innovation in DNN algorithms, and consider whether future high-performance FPGAs will outperform GPUs for next-generation DNNs. The upcoming Intel® 14-nm Stratix? 10 FPGAs will have thousands of hard floating-point units (DSPs) and on-chip RAMs (M20K memory blocks). They will also have high bandwidth memories (HBMs) and improved frequency (HyperFlex? core architecture). This combination of features brings FPGA raw floating point performance within striking distance of GPUs. Meanwhile, DNNs are quickly evolving. For example, recent innovations that exploit sparsity (e.g., pruning) and compact data types (e.g., 1-2 bit) result in major leaps in algorithmic efficiency. However, these innovations introduce irregular parallelism on custom data types, which are difficult for GPUs to handle but would be a great fit for FPGA's extreme customizability.

Interconnect power is dynamic power dissipation due to switching of interconnection capacitances. This paper describes the characterization of interconnect power in a state-of-the-art high-performance microprocessor designed for power efficiency. The analysis showed that interconnect power is over 50 % of the dynamic power. Over 90 % of the interconnect power is consumed by only 10 % of the interconnections. Relations of interconnect power to wire length distribution and hierarchy level of nets were examined. In light of the results, a router’s algorithms were modified, to use larger wire spacing and minimal length routing for the high power consuming interconnects. The power-aware router algorithm was tested on synthesized blocks, demonstrating average saving of 14 % in the dynamic power consumption without timing degradation or area increase. The results demonstrate the obtainable benefits of tuning physical design algorithms to save power.

A high-speed silicon optical modulator based on the free carrier plasma dispersion effect is presented. It is based on carrier depletion of a pn diode embedded inside a silicon-on-insulator waveguide. To achieve high-speed performance, a travelling-wave design is used to allow co-propagation of the electrical and optical signals along the length of the device. The resulting modulator has a 3 dB bandwidth of ∼30 GHz and can transmit data up to 40 Gbit/s.

The problem of predicting the next outcome of an individual binary sequence using finite memory is considered. The finite-state predictability of an infinite sequence is defined as the minimum fraction of prediction errors that can be made by any finite-state (FS) predictor. It is proven that this FS predictability can be achieved by universal sequential prediction schemes. An efficient prediction procedure based on the incremental parsing procedure of the Lempel-Ziv data compression algorithm is shown to achieve asymptotically the FS predictability. Some relations between compressibility and predictability are discussed, and the predictability is proposed as an additional measure of the complexity of a sequence.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Carrier aggregation is one of the most distinct features of 4G systems including LTEAdvanced, which is being standardized in 3GPP as part of LTE Release 10. This feature allows scalable expansion of effective bandwidth delivered to a user terminal through concurrent utilization of radio resources across multiple carriers. These carriers may be of different bandwidths, and may be in the same or different bands to provide maximum flexibility in utilizing the scarce radio spectrum available to operators. Support for this feature requires enhancement to the LTE Release 8/9 PHY, MAC, and RRC layers while ensuring that LTE Release 10 maintains backward compatibility to LTE Release 8/9. This article provides an overview of carrier aggregation use cases and the framework, and their impact on LTE Release 8/9 protocol layers.

Chip Multiprocessors (CMPs) allow different applications to concurrently execute on a single chip. When applications with differing demands for memory compete for a shared cache, the conventional LRU replacement policy can significantly degrade cache performance when the aggregate working set size is greater than the shared cache. In such cases, shared cache performance can be significantly improved by preserving the entire working set of applications that can co-exist in the cache and preserving some portion of the working set of the remaining applications.

Two CMOS differential amplifiers, one that is intended for applications in which the input common-mode range is relatively limited, the complementary self-biased differential amplifier (CSDA), and one that is intended for applications in which the input common-mode range is bounded only by the supply voltages, the very-wide-common-mode-range differential amplifier (VCDA), are discussed. Both differ from conventional CMOS differential amplifiers in having fully complementary configurations and in being self-biased through negative feedback. The amplifiers have been applied as precision high-speed comparators in commercial VLSI CMOS integrated circuits.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

volves the introduction of a new type of single player token games which are of an independent interest.

OBJECTIVE: To determine a relationship between neck circumference (NC) and risk factors for coronary heart disease by evaluating the components of the metabolic syndrome. RESEARCH METHODS AND PROCEDURES: The study group included 561 subjects (231 men and 330 women) who had no known major medical conditions and were not receiving any medication therapy. The subjects were those who attended a family health clinic for any reason between 1998 and December 2001. Main indicators studied included NC, waist circumference, waist-to-hip ratio, body mass index, blood pressure, and lipoprotein, glucose, and uric acid levels. RESULTS: Pearson's correlation coefficients indicated a significant association between NC and body mass index (men, r = 0.71; women, r = 0.81; each, p < 0.0001), waist circumference (men, r = 0.75; women, r = 0.79; each, p < 0.0001), waist-to-hip ratio (men, r = 0.56; women, r = 0.63; each, p < 0.0001), total cholesterol (men, r = 0.50; women, r = 0.66; each, p < 0.0001), low-density lipoprotein-cholesterol (men, r = 0.42; women, r = 0.60; each, p < 0.0001), triglycerides (men, r = 0.48; women, r = 0.49; each, p < 0.0001), glucose (men, r = 0.21, p < 0.001; women, r = 0.44; p < 0.0001), uric acid (men, r = 0.50, p < 0.0001; women, r = 0.60, p < 0.001), and systolic (men, r = 0.53; women, r = 0.69; each, p < 0.0001), and diastolic (men, r = 0.55; women, r = 0.65; each, p < 0.0001) blood pressure. DISCUSSION: Higher NC is correlated positively with the factors of the metabolic syndrome; therefore, it is likely to increase the risk of coronary heart disease.

Abstract In this paper, we propose a generalization of convolutional neural networks (CNN) to non‐Euclidean domains for the analysis of deformable shapes. Our construction is based on localized frequency analysis (a generalization of the windowed Fourier transform to manifolds) that is used to extract the local behavior of some dense intrinsic descriptor, roughly acting as an analogy to patches in images. The resulting local frequency representations are then passed through a bank of filters whose coefficient are determined by a learning procedure minimizing a task‐specific cost. Our approach generalizes several previous methods such as HKS, WKS, spectral CNN, and GPS embeddings. Experimental results show that the proposed approach allows learning class‐specific shape descriptors significantly outperforming recent state‐of‐the‐art methods on standard benchmarks.

In recent years, there has been tremendous interest in trying to implement the power amplifier in CMOS, due to its cost and integration benefits. Most of the high power (watt-level) CMOS PAs reported to date have not exhibited sufficient linearity required for next generation wireless standards. In this paper, we report a single-chip linear CMOS PA with sufficient power and linearity for emerging OFDM-based 4G WiMAX applications. This 90 nm 2.4 GHz CMOS linear power amplifier uses a two-stage transformer-based power combiner and produces a saturated output power of 30.1 dBm with 33% PAE and 28 dB small-signal gain. A novel bypass network is introduced to ensure stability without sacrificing gain. The choice of optimal biasing and capacitive compensation produces very flat AM-AM and AM-PM response up to high power. The PA has been tested with OFDM modulated signal and produces EVM better than -25 dB at 22.7 dBm average power. Graceful power back-off is demonstrated through turning off one of the stages, allowing low-power operation with enhanced efficiency.

We introduce an efficient computational framework for hashing data belonging to multiple modalities into a single representation space where they become mutually comparable. The proposed approach is based on a novel coupled siamese neural network architecture and allows unified treatment of intra- and inter-modality similarity learning. Unlike existing cross-modality similarity learning approaches, our hashing functions are not limited to binarized linear projections and can assume arbitrarily complex forms. We show experimentally that our method significantly outperforms state-of-the-art hashing approaches on multimedia retrieval tasks.

We report on measured radiation-induced soft error rates (SER) of memory and logic devices built in a 22 nm high-k metal gate bulk Tri-Gate technology. Our results demonstrate excellent single event upset (SEU) scaling benefits of tri-gate devices. For cosmic radiation, SEU SER reduction levels of the order of are observed relative to 32 nm planar devices, while for alpha-particles, the measured SEU SER benefit is in excess of . Similar improvements are observed for Tri-Gate combinational logic and memory array multi-cell upset (MCU) rates. Reduced SER (RSER) device SER performances (relative to standard, non -RSER devices) are on par or better than that of tested 32 nm planar devices. Finally, a novel, efficient SER reduction design called RTS is introduced.

This paper presents an image representation and matching framework for image categorization in medical image archives. Categorization enables one to determine automatically, based on the image content, the examined body region and imaging modality. It is a basic step in content-based image retrieval (CBIR) systems, the goal of which is to augment text-based search with visual information analysis. CBIR systems are currently being integrated with picture archiving and communication systems for increasing the overall search capabilities and tools available to radiologists. The proposed methodology is comprised of a continuous and probabilistic image representation scheme using Gaussian mixture modeling (GMM) along with information-theoretic image matching via the Kullback-Leibler (KL) measure. The GMM-KL framework is used for matching and categorizing X-ray images by body regions. A multidimensional feature space is used to represent the image input, including intensity, texture, and spatial information. Unsupervised clustering via the GMM is used to extract coherent regions in feature space that are then used in the matching process. A dominant characteristic of the radiological images is their poor contrast and large intensity variations. This presents a challenge to matching among the images, and is handled via an illumination-invariant representation. The GMM-KL framework is evaluated for image categorization and image retrieval on a dataset of 1500 radiological images. A classification rate of 97.5% was achieved. The classification results compare favorably with reported global and local representation schemes. Precision versus recall curves indicate a strong retrieval result as compared with other state-of-the-art retrieval techniques. Finally, category models are learned and results are presented for comparing images to learned category models.

Wearable devices can capture objective day-to-day data about Parkinson's Disease (PD). This study aims to assess the feasibility of implementing wearable technology to collect data from multiple sensors during the daily lives of PD patients. The Parkinson@home study is an observational, two-cohort (North America, NAM; The Netherlands, NL) study. To recruit participants, different strategies were used between sites. Main enrolment criteria were self-reported diagnosis of PD, possession of a smartphone and age≥18 years. Participants used the Fox Wearable Companion app on a smartwatch and smartphone for a minimum of 6 weeks (NAM) or 13 weeks (NL). Sensor-derived measures estimated information about movement. Additionally, medication intake and symptoms were collected via self-reports in the app. A total of 953 participants were included (NL: 304, NAM: 649). Enrolment rate was 88% in the NL (n = 304) and 51% (n = 649) in NAM. Overall, 84% (n = 805) of participants contributed sensor data. Participants were compliant for 68% (16.3 hours/participant/day) of the study period in NL and for 62% (14.8 hours/participant/day) in NAM. Daily accelerometer data collection decreased 23% in the NL after 13 weeks, and 27% in NAM after 6 weeks. Data contribution was not affected by demographics, clinical characteristics or attitude towards technology, but was by the platform usability score in the NL (χ2 (2) = 32.014, p<0.001), and self-reported depression in NAM (χ2(2) = 6.397, p = .04). The Parkinson@home study shows that it is feasible to collect objective data using multiple wearable sensors in PD during daily life in a large cohort.

Floating-gate MOS devices using thin tunnel oxide are becoming an acceptable standard in electrically erasable nonvolatile memory. Theoretical and experimental analysis of WRITE/ERASE characteristics for this type of memory cell are presented. A simplified device model is given based on the concept of coupling ratios. The WRITE operation is adequately represented by the simplified model. The ERASE operation is complicated due to formation of depletion layers in the transistor's channel and under the tunnel oxide. Experimental investigation of these effects is described, and they are included in a detailed cell model. In certain cell structures, a hole current can flow from the drain into the substrate during the ERASE oepration. This effect is shown to be associated with positive charge trapping in the tunnel oxide and threshold window opening. An experimental investigation of these phenomena is described, and a recommendation is made to avoid them by an appropriate cell design.

4G cellular standards are targeting aggressive spectrum reuse (frequency reuse 1) to achieve high system capacity and simplify radio network planning. The increase in system capacity comes at the expense of SINR degradation due to increased intercell interference, which severely impacts cell-edge user capacity and overall system throughput. Advanced interference management schemes are critical for achieving the required cell edge spectral efficiency targets and to provide ubiquity of user experience throughout the network. In this article we compare interference management solutions across the two main 4G standards: IEEE 802.16m (WiMAX) and 3GPP-LTE. Specifically, we address radio resource management schemes for interference mitigation, which include power control and adaptive fractional frequency reuse. Additional topics, such as interference management for multitier cellular deployments, heterogeneous architectures, and smart antenna schemes will be addressed in follow-up papers.