28-08-2017, 04:05 PM
Using past traces of traffic, we characterize network traffic over several time-of-day intervals, assuming that it is free of anomalies. We present two different approaches to characterize traffic: (i) a model-free approach based on the type method and the Sanov theorem, and (ii) a model based on the traffic modeling model using a modulated Markov process . Using these characterizations as a reference, we continuously monitor traffic and use large deviations and results from decision theory to measure the empirical measure of the traffic monitored with the corresponding reference characterization, thus identifying real-time traffic anomalies. Our experimental results show that the application of our methodology (even short-lived) anomalies are identified in a small number of observations.
The need for a space-time network naturally arises when it comes to problems such as voice recognition and time series prediction where the input signal has an explicit temporal aspect. We have demonstrated that certain tasks that do not have an explicit temporal aspect can also be processed advantageously with neural networks capable of handling temporal information.
The need for a space-time network naturally arises when it comes to problems such as voice recognition and time series prediction where the input signal has an explicit temporal aspect. We have demonstrated that certain tasks that do not have an explicit temporal aspect can also be processed advantageously with neural networks capable of handling temporal information.