This work addresses the relatively long setup latency and complicated network control management caused by on-demand virtual function service chain (vNF-SC) provisioning in inter-datacenter elastic optical networks. We first design a framework with resource pre-deployment to resolve aforementioned challenge. Specifically, is designed as discrete-time system, which operations are performed periodically fixed time slots (TS). Each TS includes phase followed phase. In phase, deep-learning (DL)...

We aim to build complex humanoid agents that integrate perception, motor control, and memory. In this work, we partly factor problem into low-level control from proprioception high-level coordination of the skills informed by vision. develop an architecture capable surprisingly flexible, task-directed a relatively high-DoF body combining pre-training controllers with high-level, task-focused controller switches among sub-policies. The resulting system is able physically-simulated solve tasks...

As artificial intelligence (AI) is playing an increasingly important role in our society and global economy, AI education literacy have become necessary components college K-12 to prepare students for AI-powered society. However, current curricula not yet been made accessible engaging enough schools from all socio-economic backgrounds with different educational goals. In this work, we developed open-source learning module high school students, which allows build their own robot companion the...

Advanced technologies are required in future mobile wireless networks to support services with highly diverse requirements terms of high data rate and reliability, low latency, massive access. Deep Learning (DL), one the most exciting developments machine learning big data, has recently shown great potential study communications. In this article, we provide a literature review on applications DL physical layer. First, analyze limitations existing signal processing techniques model accuracy,...

Semi-supervised medical image segmentation has attracted much attention in recent years because of the high cost annotations. In this paper, we propose a novel Inherent Consistent Learning (ICL) method, aims to learn robust semantic category representations through consistency guidance labeled and unlabeled data help segmentation. practice, introduce two external modules, namely Supervised Semantic Proxy Adaptor (SSPA) Unsupervised Learner (USCL) that is based on mechanism align data, as...

Nowadays, it is common for service providers (SPs) to leverage hybrid clouds improve the quality-of-service (QoS) of their Big Data applications. However, achieving guaranteed latency and/or bandwidth in its cloud, an SP might desire have a virtual datacenter (vDC) network, which can manage and manipulate network connections freely. To address this requirement, we design implement slicing orchestration (NSO) system that create expand vDCs across optical/packet domains on-demand. Considering...

As Deep learning (DL) systems continuously evolve and grow, assuring their quality becomes an important yet challenging task. Compared to non-DL systems, DL have more complex team compositions heavier data dependency. These inherent characteristics would potentially cause be vulnerable bugs and, in the long run, maintenance issues. Code smells are empirically tested as efficient indicators of systems. Therefore, we took a step forward into identifying code smells, understanding impact on...

As artificial intelligence (AI) is playing an increasingly important role in our society and global economy, AI education literacy have become necessary components college K-12 to prepare students for AI-powered society. However, current curricula not yet been made accessible engaging enough schools from all socio-economic backgrounds with different educational goals. In this work, we developed open-source learning module high school students, which allows build their own robot companion the...

We use over 350,000 Yelp reviews on 5,000 restaurants to perform an ablation study text preprocessing techniques. also compare the effectiveness of several machine learning and deep models predicting user sentiment (negative, neutral, or positive). For models, we find that using binary bag-of-word representation, adding bi-grams, imposing minimum frequency constraints normalizing texts have positive effects model performance. pre-trained word embeddings capping maximum length often boost...

The carrier frequency offset (CFO) estimation is a crucial problem in orthogonal division multiplexing (OFDM) systems, especially for the enabling technologies Internet of Things, which demand stringent limitations low cost, power consumption, and wide ranges, like wireless sensor networks. In this letter, we propose an efficient CFO method OFDM systems with in- quadrature-phase (IQ) imbalance, combines emerging multitask learning (MTL) channel residual energy (CRE) to reduce average...

Purpose For start-ups or growing firms, to effectively navigate the unpredictable nature of digital development and achieve superior innovative performance, it is crucial have a workforce comprised creative employees. Drawing upon principles social information processing theory, this study aims investigate whether specific combinations organizational internal external environments, as well work characteristics in age, can foster high level employee behavior. Design/methodology/approach By...

In view of that the reliability existing typical IP city-level geolocation approaches are limited to delay measurement accuracy, targets reachability and other factors, this paper proposed an approach based on PoP-level (Point Presence) network analysis. First, "low-high-low" distribution characteristics one-hop-delay among nodes in different cities revealed. The connections targets' obtained by partitioning probe paths one-hop-delay. Secondly, these analyzed extracting closely connected...

Software-defined elastic optical networks (SD-EONs) provide operators more flexibility to customize their infrastructure dynamically and adaptively, network virtualization, i.e., infrastructure-as-a-service (IaaS), enables multiple tenants share the substrate efficiently. In this paper, we study how provision virtual SD-EONs (vSD-EONs) with correlated data control plane embedding (χ-VNE) that considers quality-of-service (QoS) of (vCP), availability channel latency. We propose a χ-VNE...

Today's networked and distributed applications, by large, rely on cloud services. However, solely cloud-based services are not the ideal solution for all use cases, in particular, case of high volume data sharing scientific computing whose usage costs could be prohibitively high. Thus we take a task building distributed, federated repository, dubbed Hydra, large data. In this paper, compare two design choices: designing Hydra over TCP/IP with centralized controller, Named Data Network (NDN)...

Due to the technical and cost limitations, wireless systems suffer from various hardware impairments, including phase noise, power amplifier nonlinearity, carrier frequency offset in-phase quadrature-phase imbalance. These impairments can highly degrade physical layer performance are usually compensated separately by using model-based signal processing techniques. However, due high large bandwidth of 5G new radio, coupling effects between different aggravated, which greatly degrades...

Orthogonal frequency-division multiplexing (OFDM) is widely adopted in narrowband Internet of Things (NB-IoT). Nevertheless, the OFDM system highly sensitive to impairments caused by imperfect radio-frequency hardwares, which may greatly jeopardize orthogonality between different subcarriers and degrade demodulation performance. Though large efforts have been devoted hardware impairment estimation, however, it challenging jointly estimate multiple due their coupling effects, especially for...

Microsoft's internal big data analytics platform is comprised of hundreds thousands machines, serving over half a million jobs daily, from users. The majority these are recurring and crucial for the company's operation. Although administrators spend significant effort tuning system performance, some inevitably experience slowdowns, i.e., their execution time degrades previous runs. Currently, investigation such slowdowns labor-intensive error-prone process, which costs Microsoft human...