INTRODUCTIONWith the worldwide availability of a large swath of spectrum at the 60 GHz band for unlicensed use, we are starting to see an emergence of new technologies enabling Wi-Fi communication in this frequency band. However, signal propagation at the 60 GHz band significantly differs from that at the 2.4 and 5 GHz bands. Therefore, efficient use of this vast spectrum resource requires a fundamental rethinking of the operation of Wi-Fi and a transition from omnidirectional to directional wireless medium usage. The IEEE 802.11ad amendment addresses these challenges, bringing multi-gigabit-per-second throughput and new application scenarios to Wi-Fi users. These new uses include instant wireless synchronization, high-speed media file exchange between mobile devices without fixed network infrastructure, and wireless cable replacement (e.g., to connect to high definition wireless displays).The most significant difference in 60 GHz propagation behavior is increased signal attenuation. At a typical IEEE 802.11ad range of 10 m, additional attenuation of 22 dB compared to the 5 GHz band is predicted by the Friis transmission equation, resulting from the frequencydependent difference in antenna aperture. In contrast, oxygen absorption plays a minor role over short-range distances, even though it peaks at 60 GHz [1]. Furthermore, 60 GHz communication is characterized by a quasi-optical propagation behavior [2] where the received signal is dominated by the line of sight (LOS) path and first order reflections from strong reflecting materials. As an example, metallic surfaces were found to be strong reflectors and allow non-LOS (NLOS) communication [2]. Concrete materials, on the other hand, cause additional large signal attenuation and can easily create a blockage. Thus, 60 GHz communication is more suitable to in-room environments where sufficient reflectors are present.This article discusses the design assumption resulting from the millimeter-wave (mm-Wave) propagation characteristics and related adaptation to the 802.11 architecture. We further present typical device configurations, an overview of the IEEE 802.11ad physical (PHY) layer, and the newly introduced personal basic service set network architecture. This is followed by an in-depth description of the IEEE 802.11ad beamforming (BF) mechanism and hybrid medium access control (MAC) design, which are the central elements to facilitate directional communication. DIRECTIONAL COMMUNICATIONThe IEEE 802.11ad amendment to the 802.11 standard defines a directional communication scheme that takes advantage of beamforming antenna gain to cope with increased attenuation in the 60 GHz band [1]. With quasi-optical propagation behavior, low reflectivity, and high attenuation, beamforming results in a highly directional signal focus. Based on this behavior, the standard introduces a novel concept of "virtual" antenna sectors [3] that discretize the antenna azimuth. IEEE 802.11ad sectors can be implemented either using precomputed antenna weight vectors for a phased antenna arr...

No abstract

No abstract

No abstract

In this paper, we demonstrate the efficacy of transfer learning and continuous learning for various automatic speech recognition (ASR) tasks using end-to-end models trained with CTC loss. We start with a large pre-trained English ASR model and show that transfer learning can be effectively and easily performed on: (1) different English accents, (2) different languages (from English to German, Spanish, Russian, or from Mandarin to Cantonese) and (3) application-specific domains. Our extensive set of experiments demonstrate that in all three cases, transfer learning from a good base model has higher accuracy than a model trained from scratch. Our results indicate that, for fine-tuning, larger pre-trained models are better than small pre-trained models, even if the dataset for fine-tuning is small. We also show that transfer learning significantly speeds up convergence, which could result in significant cost savings when training with large datasets.

No abstract

No abstract

Congestion in interconnection networks due to the presence of hot spo t s i s a n i m p ortant and di cult problem that occurs in parallel machines. This problem has been studied in depth and di erent solutions for the case of multiprocessors with shared memory have been proposed. Current trends point towards the implementation of systems with physically distributed memory, either based on message passing (multicomputers) or on a single shared memory address space (multiprocessors). Our paper is developed i n t h i s c ontext. Up to now, proposals to improve the throughput of networks with hot-spots have focused on using virtual channels or adaptive algorithms. We present a novel solution based on recon gurable networks. A recon gurable network is one in which nodes can change their position depending on the communication pattern in order to diminish the congestion produced in the network and, therefore, increase its throughput. We studied this problem in two-dimensional k-ary n-cube networks using a deterministic routing algorithm and wormhole routing. In this paper the main features of a recon gurable network are p r esented and the results obtained by simulation are shown. These results con rm that this technique as a very interesting one for systems with distributed memory, with applications to a great variety of problems.