Project title: Recurrent Deep Learning Machine for Robust, Adaptive, or Accommodative Filtering
Dr. James Lo has been awarded a $340,736, three-year grant (ECCS-1508880) for the period 2015–18 from the National Science Foundation to develop robust and adaptive filters through a neural network approach. The award includes funds to support a Graduate Research Assistant for the duration of the grant.
The prediction and estimation of a process, called a signal process, given a relevant process, called measurement process, both of which may involve randomness, is a fundamental problem in a large range of fields. As the signal evolves and measurements keep coming in, an algorithm is needed to predict or estimate the signal and update the best summary of the information extracted from the measurements up to each time point using the measurement at the same time point. Such a recursive algorithm is called a filter.
When the signal or measurement process is affected by an uncertain or changing environment, a filter that adapts to the environment is called an adaptive filter.
In many applications, whether or not an uncertain or changing environment is involved, a large individual error in estimation or prediction may cause an undesirable consequence, a filter that can reduce large errors is called a robust filter. A robust filter must balance filtering accuracy and robustness in a way required by the application.
A neural network approach for optimal filtering of nonlinear signals/measurements was introduced by Dr. Lo in 1992. The purpose of the present project supported by the subject grant is to extend the approach and develop robust and adaptive filters.
Dr. James Lo has been awarded a $340,736, three-year grant (ECCS-1508880) for the period 2015–18 from the National Science Foundation to develop robust and adaptive filters through a neural network approach. The award includes funds to support a Graduate Research Assistant for the duration of the grant.
The prediction and estimation of a process, called a signal process, given a relevant process, called measurement process, both of which may involve randomness, is a fundamental problem in a large range of fields. As the signal evolves and measurements keep coming in, an algorithm is needed to predict or estimate the signal and update the best summary of the information extracted from the measurements up to each time point using the measurement at the same time point. Such a recursive algorithm is called a filter.
When the signal or measurement process is affected by an uncertain or changing environment, a filter that adapts to the environment is called an adaptive filter.
In many applications, whether or not an uncertain or changing environment is involved, a large individual error in estimation or prediction may cause an undesirable consequence, a filter that can reduce large errors is called a robust filter. A robust filter must balance filtering accuracy and robustness in a way required by the application.
A neural network approach for optimal filtering of nonlinear signals/measurements was introduced by Dr. Lo in 1992. The purpose of the present project supported by the subject grant is to extend the approach and develop robust and adaptive filters.