Abstract Minimizing delay and improving efficiency in content delivery is essential in resource hungry networks. Caching enables such networks like Mobile Ad Hoc Network (MANET) to have improved performance as cached data is shared with coordination. It also helps nodes to operate offline with the help of cached content besides promoting availability and accessibility. Caching is the process of storing repeatedly... Minimizing delay and improving efficiency in content delivery is essential in resource hungry networks. Caching enables such networks like Mobile Ad Hoc Network (MANET) to have improved performance as cached data is shared with coordination. It also helps nodes to operate offline with the help of cached content besides promoting availability and accessibility. Caching is the process of storing repeatedly used data in a temporary memory thus avoiding unnecessary processing which leads to waste of time and resources. However, it is crucial to know how data is to be cached, how long it has to be retained and when it has to be replaced with new content as it has finite size. Cache replacement is an important activity which needs intelligence for efficiency. In this paper we proposed an enhanced POC tree based algorithm for association rule mining. The association rules of the proposed algorithm provide significant heuristics to make cache replacement decisions besides achieving correlation of data items. This algorithm employs light-weight and scalable data structure known as Nodesets. We compared the proposed approach with existing algorithms. Our simulation results found the proposed method outperforms existing ones. Read More Read Less

Abstract Computer Vision leads to large applications for image and video classification and segmentation. It provides the increasing focus of the community on this topic, has generated awide scope of approaches that use different kernel shapes and techniques for executing convolutions with respect to the classic one, such as for example separable convolutions, deformable convolutions or deconvolutions ([4, 5]), frequently used in... Computer Vision leads to large applications for image and video classification and segmentation. It provides the increasing focus of the community on this topic, has generated awide scope of approaches that use different kernel shapes and techniques for executing convolutions with respect to the classic one, such as for example separable convolutions, deformable convolutions or deconvolutions ([4, 5]), frequently used in semantic segmentation tasks ([23, 13]). It has been found that it is common knowledge that FPGAs can be used to accelerate classic Convolutional layers in CNNs, there is limited literature about FPGA-based accelerators supporting less regular and common processing kernels ([20]). This work begins from the previous experience acquired developing NEURAghe, we try to improve exibility of CNN accelerators and to study new methodologies to improve efficiency on the previously mentioned use-cases. Initially we focus on layered approaches based on 1D convolutions, that, as indicated by several recent research results, can be effectively used to classify and segment time series and sequences, as well as in tasks involving sequence modeling. In multiple scenarios a convolution approach applied on the time dimension, hereafter called Temporal Convolution Network (TCN) can outperform classic strategies relying on recurrent networks in terms of accuracy and training time. We modified NEURAghe to support TCN and validate results on an ECG-classification benchmark, achieving up to 95% efficiency in terms of GOPS/s with respect to the accelerator peak performance. Read More Read Less

Abstract Intent detection and slot filling are the two major tasks of the basis of dialogue system and human -computer interaction. Normally these are performed in a pipeline fashion, suffering from error propagation. IN order to address this issue, this work advocates a method to solve intent detection and slot filling jointly by back introducing the result of slot filling into... Intent detection and slot filling are the two major tasks of the basis of dialogue system and human -computer interaction. Normally these are performed in a pipeline fashion, suffering from error propagation. IN order to address this issue, this work advocates a method to solve intent detection and slot filling jointly by back introducing the result of slot filling into intent detection step. Besides, rules extracted on the training set are used to ease the noise and imbalance problems. Experimental outcomes are generated by fusing the results of rule-based method and model-based method. Finally it shows that our method is effective and achieves best performance among all teams. Read More Read Less