Learning Unitary Operators with Help From u(n)
This work addresses the vanishing/exploding gradient problem in recurrent neural networks, offering a practical parametrization for unitary operators, though it is incremental as it builds on existing unitary RNN methods.
The paper tackles the challenge of parametrizing unitary operators to address vanishing/exploding gradients in recurrent neural networks by using the Lie algebra u(n) to define a simple gradient descent space with n^2 real coefficients. It demonstrates effectiveness by outperforming baselines in learning arbitrary unitary operators and generalizing a unitary RNN to solve long-memory tasks.
A major challenge in the training of recurrent neural networks is the so-called vanishing or exploding gradient problem. The use of a norm-preserving transition operator can address this issue, but parametrization is challenging. In this work we focus on unitary operators and describe a parametrization using the Lie algebra $\mathfrak{u}(n)$ associated with the Lie group $U(n)$ of $n \times n$ unitary matrices. The exponential map provides a correspondence between these spaces, and allows us to define a unitary matrix using $n^2$ real coefficients relative to a basis of the Lie algebra. The parametrization is closed under additive updates of these coefficients, and thus provides a simple space in which to do gradient descent. We demonstrate the effectiveness of this parametrization on the problem of learning arbitrary unitary operators, comparing to several baselines and outperforming a recently-proposed lower-dimensional parametrization. We additionally use our parametrization to generalize a recently-proposed unitary recurrent neural network to arbitrary unitary matrices, using it to solve standard long-memory tasks.