Deep Learning Model Reuse and Composition in Knowledge Centric Networking

Sapra, Dolly; Pimentel, Andy D.

doi:10.1109/icccn49398.2020.9209668

Cited by 6 publications

(3 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ML models that are capable of generating knowledge are usually trained by vendors, who may keep the models proprietary. However, keeping them proprietary can cause redundant computations and an inefficient use of resources [199]. ML or heuristic models can be used to detect features from input information/data to create knowledge.…”

Section: Generating Knowledge Using Machine Learning Methodsmentioning

confidence: 99%

A Comprehensive Survey on Knowledge-Defined Networking

Wijesekara

Gunawardena

2023

Telecom

View full text Add to dashboard Cite

Traditional networking is hardware-based, having the control plane coupled with the data plane. Software-Defined Networking (SDN), which has a logically centralized control plane, has been introduced to increase the programmability and flexibility of networks. Knowledge-Defined Networking (KDN) is an advanced version of SDN that takes one step forward by decoupling the management plane from control logic and introducing a new plane, called a knowledge plane, decoupled from control logic for generating knowledge based on data collected from the network. KDN is the next-generation architecture for self-learning, self-organizing, and self-evolving networks with high automation and intelligence. Even though KDN was introduced about two decades ago, it had not gained much attention among researchers until recently. The reasons for delayed recognition could be due to the technology gap and difficulty in direct transformation from traditional networks to KDN. Communication networks around the globe have already begun to transform from SDNs into KDNs. Machine learning models are typically used to generate knowledge using the data collected from network devices and sensors, where the generated knowledge may be further composed to create knowledge ontologies that can be used in generating rules, where rules and/or knowledge can be provided to the control, management, and application planes for use in decision-making processes, for network monitoring and configuration, and for dynamic adjustment of network policies, respectively. Among the numerous advantages that KDN brings compared to SDN, enhanced automation and intelligence, higher flexibility, and improved security stand tall. However, KDN also has a set of challenges, such as reliance on large quantities of high-quality data, difficulty in integration with legacy networks, the high cost of upgrading to KDN, etc. In this survey, we first present an overview of the KDN architecture and then discuss each plane of the KDN in detail, such as sub-planes and interfaces, functions of each plane, existing standards and protocols, different models of the planes, etc., with respect to examples from the existing literature. Existing works are qualitatively reviewed and assessed by grouping them into categories and assessing the individual performance of the literature where possible. We further compare and contrast traditional networks and SDN against KDN. Finally, we discuss the benefits, challenges, design guidelines, and ongoing research of KDNs. Design guidelines and recommendations are provided so that identified challenges can be mitigated. Therefore, this survey is a comprehensive review of architecture, operation, applications, and existing works of knowledge-defined networks.

show abstract

Section: Generating Knowledge Using Machine Learning Methodsmentioning

confidence: 99%

A Comprehensive Survey on Knowledge-Defined Networking

Wijesekara

Gunawardena

2023

Telecom

View full text Add to dashboard Cite

show abstract

“…Intelligence can be algorithms that aggregate information into intelligent fragments, i.e., intelligence models, or abstract information obtained through application models, i.e., intelligence samples [6]. For example, a trained ML model is a kind of intelligence that can return synthetic intelligence from input information [41]. The intelligence extracted by advanced technologies can be further utilized by the intelligence combination method, in which existing intelligence is further analyzed/organized to generate new intelligence.…”

Section: Motivations Of the Internet Of Intelligencementioning

confidence: 99%

Internet of Intelligence: A Survey on the Enabling Technologies, Applications, and Challenges

Tang,

Yu,

Xie

et al. 2022

Preprint

View full text Add to dashboard Cite

The Internet of intelligence is conceived as an emerging networking paradigm, which will make intelligence as easy to obtain as information. This paper provides an overview of the Internet of intelligence, focusing on motivations, architecture, enabling technologies, applications, and existing challenges. This can provide a good foundation for those who are interested to gain insights into the concept of the Internet of intelligence and the key enablers of this emerging networking paradigm. Specifically, this paper starts by investigating the evolution of networking paradigms and artificial intelligence (AI), based on which we present the motivations of the Internet of intelligence by demonstrating that networking needs intelligence and intelligence needs networking. We then present the layered architecture to characterize the Internet of intelligence systems and discuss the enabling technologies of each layer. Moreover, we discuss the critical applications and their integration with the Internet of intelligence paradigm. Finally, some technical challenges and open issues are summarized to fully exploit the benefits of the Internet of intelligence.

show abstract

“…The deployment of neural networks in a wide range of settings requires a multi-objective perspective these days with the advent of connected and intelligent edge devices [42,56]. These devices do not have large computation or storage capabilities.…”

Section: Related Workmentioning

confidence: 99%

Designing convolutional neural networks with constrained evolutionary piecemeal training

Sapra

Pimentel

2021

Appl Intell

Self Cite

View full text Add to dashboard Cite

The automated architecture search methodology for neural networks is known as Neural Architecture Search (NAS). In recent times, Convolutional Neural Networks (CNNs) designed through NAS methodologies have achieved very high performance in several fields, for instance image classification and natural language processing. Our work is in the same domain of NAS, where we traverse the search space of neural network architectures with the help of an evolutionary algorithm which has been augmented with a novel approach of piecemeal-training. In contrast to the previously published NAS techniques, wherein the training with given data is considered an isolated task to estimate the performance of neural networks, our work demonstrates that a neural network architecture and the related weights can be jointly learned by combining concepts of the traditional training process and evolutionary architecture search in a single algorithm. The consolidation has been realised by breaking down the conventional training technique into smaller slices and collating them together with an integrated evolutionary architecture search algorithm. The constraints on architecture search space are placed by limiting its various parameters within a specified range of values, consequently regulating the neural network’s size and memory requirements. We validate this concept on two vastly different datasets, namely, the CIFAR-10 dataset in the domain of image classification, and PAMAP2 dataset in the Human Activity Recognition (HAR) domain. Starting from randomly initialized and untrained CNNs, the algorithm discovers models with competent architectures, which after complete training, reach an accuracy of of 92.5% for CIFAR-10 and 94.36% PAMAP2. We further extend the algorithm to include an additional conflicting search objective: the number of parameters of the neural network. Our multi-objective algorithm produces a Pareto optimal set of neural networks, by optimizing the search for both the accuracy and the parameter count, thus emphasizing the versatility of our approach.

show abstract

Deep Learning Model Reuse and Composition in Knowledge Centric Networking

Cited by 6 publications

References 35 publications

A Comprehensive Survey on Knowledge-Defined Networking

A Comprehensive Survey on Knowledge-Defined Networking

Internet of Intelligence: A Survey on the Enabling Technologies, Applications, and Challenges

Designing convolutional neural networks with constrained evolutionary piecemeal training

Contact Info

Product

Resources

About