AMATH Grad Students
Avneet Kaur | University of Waterloo
System estimation using a recurrent neural network
While it is well-known that a Kalman filter is an optimal state estimator of linear systems (given Gaussian noises), optimal estimation of non-linear systems of ordinary differential equations is an open problem. In this talk we will discuss some first steps on a pathway to optimal state estimation using neural nets. We propose to use a recurrent neural network(RNN) to estimate the state of a system. Our codes will be tested on a number of examples, including the Lorenz attractor. The performance of the neural net based estimator will be compared to the popular extended Kalman filter(EKF).