Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems.

/pdf/xgboost-a-scalable-tree-boosting-system-48ocu6x4c7.pdf

XGBoost: A Scalable Tree Boosting System

This open source computing framework unifies streaming, batch, and interactive big data workloads to unlock new applications

https://dl.acm.org/doi/pdf/10.1145/2934664

Apache Spark: a unified engine for big data processing

The scaling of microchip technologies has enabled large scale systems-on-chip (SoC). Network-on-chip (NoC) research addresses global communication in SoC, involving (i) a move from computation-centric to communication-centric design and (ii) the implementation of scalable communication structures. This survey presents a perspective on existing NoC research. We define the following abstractions: system, network adapter, network, and link to explain and structure the fundamental concepts. First, research relating to the actual network design is reviewed. Then system level design and modeling are discussed. We also evaluate performance analysis techniques. The research shows that NoC constitutes a unification of current trends of intrachip communication rather than an explicit new alternative.

/pdf/a-survey-of-research-and-practices-of-network-on-chip-2rc0mlsrfq.pdf

A survey of research and practices of Network-on-chip

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems-patient classification, fundamental biological processes and treatment of patients-and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine.

https://royalsocietypublishing.org/doi/pdf/10.1098/rsif.2017.0387

Opportunities and obstacles for deep learning in biology and medicine.

Receivers and reception methods conduct spread spectrum reception using the energy of multiple bits, preferably a substantial number of bits, in a packet for correct alignment of the spreading code at the receiver. Code acquisition, as well as packet detection and acquisition are provided by embodiments of the invention. Preferred embodiment receivers and reception methods provide packet detection at low SINR looking at energy of multiple bits of the packet, frame synchronization using short preamble, timing recovery at low SINR looking at energy of multiple bits of the packet, quick code acquisition using parallel search for all possible code phases, code acquisition at low SINR, and tracking of acquired timing and code phase. Preferred embodiment receivers and reception methods use the energy of multiple bits for making decisions.

Spread-spectrum receiver and reception method

A low signal to interference and noise ratio (SINR) wireless environment makes successful reception and accurate detection of packets a challenging task. We can combat a low SINR by using direct-sequence spread spectrum modulation with a sufficiently high processing gain, as well as by employing block codes for FEC on each packet. In this paper, we propose a packet detection and acquisition method for low SINR environments. The received waveform is processed using a non-linear filtering algorithm based on a maximum likelihood ratio calculation to detect the presence of transmitted packets. The algorithm uses the entire packet transmission for chip timing synchronization and code acquisition at the receiver. The performance of the algorithm is analyzed using simula- tions and real packet transmissions captured over the wireless medium. The algorithm is finally implemented on a FPGA which is a part of a wireless transceiver prototype. The experimental results are reported.

Packet detection and acquisition at low SINR in spread-spectrum based wireless communications

We consider the problem of code acquisition at low signal to interference and noise ratio (SINR) in packet-based spread spectrum communications, where a centralized ?pilot? synchronization signal and high-precision oscillators are not available. Such problems arise in the context of ad-hoc networks, for example, which require fast code acquisition times and gradual degradation of the network with decreasing SINR. We motivate the use of the ?postdetection integration? approach, which utilizes the energy of multiple bits in a packet, for code acquisition at low SINR. We present the implementation of a fully parallel architecture which simultaneously looks at all possible code alignments over multiple bits. This leads to a drastic reduction in acquisition time compared to serial search based methods. We report some preliminary simulation and experimental results from a hardware prototype of a transceiver on which the code acquisition algorithm was implemented.

Code Acquisition at Low SINR in Spread Spectrum Communications

In accordance with embodiments disclosed herein, there are provided apparatus, systems and methods for optimizing DSL line configurations and parameters in a DSL system. For example, such a system may include means for: interfacing to a currently operating Digital Subscriber Line ("DSL line") operating within a DSL system having a plurality of DSL lines operating therein; evaluating an active configuration of the currently operating DSL line by collecting operational data for the DSL line during operation of the DSL line; determining a plurality of candidate configurations for the currently operating DSL line, the plurality of candidate configurations being operationally compatible with the currently operating DSL line; selecting one of the sub-set of the plurality of candidate configurations as the selected configuration for the DSL line based on a preferred operational characteristic; and transitioning the currently operating DSL line from operating using the active configuration for the DSL line to operating using the selected configuration for the DSL line. Other related embodiments are disclosed.

Systems, methods, and apparatuses for optimizing dsl line configurations and parameters in a dsl system

Currently deployed asynchronous wireless networks, e.g., 802.11a/b/g, have a limited spatial range and low reliability due to their inability to operate at low signal to interference and noise ratio (SINR). Low SINR operation can be achieved using a direct-sequence spread-spectrum (DSSS) based communication with a high spreading gain. The reception at low SINR is limited by the inability to synchronize at a low SINR. In this paper, we present the design of a synchronization block which enables a receiver to synchronize to an incoming packet at a low SINR with a high reliability. It leverages an algorithm for code acquisition in asynchronous packet-based reception described in [1]. The design of this block also consists of novel algorithms for code tracking, frame synchronization and multipath search for RAKE combining. A payload based on binary orthogonal signaling is attached to the packet header to complete an asynchronous packet-based system. The simulation results show that each of the synchronization algorithms as well as the entire synchronization block operate reliably at low SINR at the input of the receiver.

Manish Amde

Papers

Spread-spectrum receiver and reception method

Packet detection and acquisition at low SINR in spread-spectrum based wireless communications

Code Acquisition at Low SINR in Spread Spectrum Communications

Systems, methods, and apparatuses for optimizing dsl line configurations and parameters in a dsl system

Synchronization at Low SINR in Asynchronous Direct-Sequence Spread-Spectrum Communications