Annealed Competition of Experts for a Segmentation and Classiication of Switching Dynamics
نویسنده
چکیده
We present a method for the unsupervised segmentation of data streams originating from diierent unknown sources which alternate in time. We use an architecture consisting of competing neural networks. Memory is included in order to resolve ambiguities of input-output relations. In order to obtain maximal specialization , the competition is adiabatically increased during training. Our method achieves almost perfect identiication and segmentation in the case of switching chaotic dynamics where input manifolds overlap and input-output relations are ambiguous. Only a small dataset is needed for the training proceedure. Applications to time series from complex systems demonstrate the potential relevance of our approach for time series analysis and short-term prediction.
منابع مشابه
Annealed Competition of Experts for a Segmentation and Classification of Switching Dynamics
متن کامل
Segmentation of switching dynamics with a Hidden Markov Model of neural prediction experts
We discuss a framework for modeling the switching dynamics of a time series based on hidden Markov models (HMM) of prediction experts, here neural networks. Learning is treated as a maximum likelihood problem. In particular, we present an Expectation-Maximization (EM) algorithm for adjusting the expert parameters as well as the HMM transition probabilities. Based on this algorithm, we develop a...
متن کاملDeveloping a Macro-segmentation Model at Industry Level: Iranian Banking Industry
Drastic changes and turbulence in macro-economic factors have the greatest impact on banks target market attractiveness in Iran. It is assumed that conventional segmentation models at the corporate level are not efficient for banking system. This study aims to develop a new segmentation model at the industry level for banks of Iran. For this purpose, structures and variables at the industry lev...
متن کاملDynamics of an eco-epidemic model with stage structure for predator
The predator-prey model with stage structure for predator is generalized in the context of ecoepidemiology, where the prey population is infected by a microparasite and the predator completely avoids consuming the infected prey. The intraspecific competition of infected prey is considered. All the equilibria are characterized and the existence of a Hopf bifurcation at the coexistence equilibriu...
متن کاملDynamic segmentation and ranking approach of customers and identifying their behavioral mobility using data mining techniques in Kargaran Welfare Bank
Nowadays, identifying, determining the value and segmentation of customers is essential for a bank. Dynamic classification of workers' welfare bank customers and identification of their behavioral mobility between different departments in a specific period of time using data techniques Kaveh. In this regard, transaction data of customers of this bank was considered as a statistical community. I...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 1996