1 / 10

Introduction to Machine Learning, its potential usage in network area, & the proposed NMLRG

Introduction to Machine Learning, its potential usage in network area, & the proposed NMLRG Proposed NMLRG IETF 94, November 2015. draft-jiang-nmlrg-network-machine-learning Sheng Jiang (Speaker). Page 1 / 10. Motivation.

Download Presentation

Introduction to Machine Learning, its potential usage in network area, & the proposed NMLRG

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Introduction to Machine Learning,its potential usage in network area, & the proposed NMLRG Proposed NMLRGIETF 94, November 2015 draft-jiang-nmlrg-network-machine-learning Sheng Jiang (Speaker) Page 1/10

  2. Motivation • Networks and network problems become more and more complicated, many varieties and dynamically changing • Looking for new mechanism that can adapt to various and dynamic environment • Looking for autonomic mechanism to replace human operations, even human programming • Machine learning was also motivated by tasks that are extremely difficult to program by hand • Advantages: robustly solve complicated tasks, reliance on real-world data instead of pure intuition, be able to adapt to new situations Page 2/10

  3. Machine Learning Traditional Future Approximate data that describe the task Accurate data that describe the task (match the pre-designed conditions) • Machine learning technologies can learn from historical data, and make predictions or decisions, rather than following strictly static program instructions • They can dynamically adapt to a changing situation and enhance their own intelligence by learning from new data • They have been successful in image analysis, language recognition, conversation simulation, and many other applications Designed machine learning patterns and algorithms Target Result Computing Process Computing Process if-elsePrograms (Fixed Human Knowledge) LearnedRules Target Result Computing Process Data Page 3/10

  4. Machine Learning Categories • Machine learning mechanisms are typically classified into three broad categories, depending on the nature of the learning “signal” or “feedback” available: supervised learning, unsupervised learning, reinforcement learning. Also semi-supervised learning • There are a few basic machine learning approaches. They can be mixed together to complete complicated tasks • Classification, clustering, regression, anomaly detection, density estimation, dimensionality reduction, decision tree learning, association rule learning, artificial neural networks, reinforcement learning, similarity and metric learning, representation learning, etc. Page 4/10

  5. Precondition of Applying Machine Learning Approach • Although it is different from big data or data mining, machine learning does also need data. However, machine learning can be applied with small set of data or dynamic feedback from environment. The quality of data decides the efficient and accuracy of machine learning result • There is no generic machine learning mechanism that could suitable for all or most of use cases. For each use case, the developers need to design a specific learning path, which may combine multiple approaches or algorithms together. The feature design and learning path design are the key factors in the machine learning applications Page 5/10

  6. Limitation of Machine Learning Mechanism • So far, the machine learning mechanism does not perform very well for accurate result. In most successful cases, it is used as an assistant analysis tool. Its results are usually accepted in fault- tolerant environment or with further human confirmation • Quality requirements on data • Most important elements must be measurable or quantifiable Page 6/10

  7. More Considerations • To achieve autonomic decision or minimize the human intervention, there should be evaluation system for the results of machine learning mechanism. The evaluation system and machine learning mechanism could compose a close decision loop for autonomic decision • Priori knowledge vs. universal • For many use cases, human interactive is very important to improve the usability of the result • Combining with traditional mechanisms • Incremental mining • Self iterative optimization Page 7/10

  8. Potential Usage in Network Area • The machine learning mechanism can be used to intelligently learn the various environments of networks and react to dynamic situations • Many network aspect can benefit: network establishing, controlling, managing, network applications and customer services, etc. • acquire knowledge from the existing networks so that new networks can be established with minimum efforts; • use machine learning mechanisms for routing control and optimization; • predict future network status in network management; • autonomic and dynamically manage the network; • analyze network faults and support recovery; • learn network attacks and their behaviors, so that protection mechanisms could be self-developed; • unify the data structure and the communication interface between network/network devices and customers, so that the upper-layer applications could easily obtain relevant network information, etc. Page 8/10

  9. Proposed NMLRG • The Network Machine Learning Research Group (NMLRG) provides a forum for researchers to explore the potential of machine learning technologies for networks. In particular, the NMLRG will work on potential approaches that apply machine learning technologies in network control, network management, and supplying network data for upper-layer applications • The initial focus of the NMLRG will be on higher-layer concepts, use cases • The NMLRG is expected to identify and document requirements, to survey possible approaches, to provide specifications for proposed solutions, and to prove concepts with prototype implementations that can be tested in real-world environments Page 9/10

  10. Comments & discussionare welcomed! Thank You! jiangsheng@huawei.com Page 10/10

More Related