English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Online learning (OL) from data streams is an emerging area of research thatencompasses numerous challenges from stream processing, machine learning, andnetworking. Recent extensions of stream-processing platforms, such as ApacheKafka and Flink, already provide basic extensions for the training of neuralnetworks in a stream-processing pipeline. However, these extensions are notscalable and flexible enough for many real-world use-cases, since they do notintegrate the neural-network libraries as a first-class citizen into theirarchitectures. In this paper, we present TensAIR, which provides an end-to-enddataflow engine for OL from data streams via a protocol to which we refer asasynchronous iterative routing. TensAIR supports the common dataflow operators,such as Map, Reduce, Join, and has been augmented by the data-parallel OLfunctions train and predict. These belong to the new Model operator, in whichan initial TensorFlow model (either freshly initialized or pre-trained) isreplicated among multiple decentralized worker nodes. Our decentralizedarchitecture allows TensAIR to efficiently shard incoming data batches acrossthe distributed model replicas, which in turn trigger the model updates viaasynchronous stochastic gradient descent. We empirically demonstrate thatTensAIR achieves a nearly linear scale-out in terms of (1) the number of workernodes deployed in the network, and (2) the throughput at which the data batchesarrive at the dataflow operators. We exemplify the versatility of TensAIR byinvestigating both sparse (Word2Vec) and dense (CIFAR-10) use-cases, for whichwe are able to demonstrate very significant performance improvements incomparison to Kafka, Flink, and Horovod. We also demonstrate the magnitude ofthese improvements by depicting the possibility of real-time concept driftadaptation of a sentiment analysis model trained over a Twitter stream.

TensAIR: Online Learning from Data Streams via Asynchronous Iterative Routing