VPNet: Variable Projection Networks
This work addresses efficiency challenges in signal processing tasks like classification, regression, and clustering, though it appears incremental as it builds on existing variable projection methods.
The authors tackled the problem of designing efficient neural networks for signal processing by introducing VPNet, a model-driven architecture based on variable projection, which achieved good accuracy and fast learning at low computational cost compared to fully connected and convolutional networks.
We introduce VPNet, a novel model-driven neural network architecture based on variable projection (VP). Applying VP operators to neural networks results in learnable features, interpretable parameters, and compact network structures. This paper discusses the motivation and mathematical background of VPNet and presents experiments. The VPNet approach was evaluated in the context of signal processing, where we classified a synthetic dataset and real electrocardiogram (ECG) signals. Compared to fully connected and one-dimensional convolutional networks, VPNet offers fast learning ability and good accuracy at a low computational cost of both training and inference. Based on these advantages and the promising results obtained, we anticipate a profound impact on the broader field of signal processing, in particular on classification, regression and clustering problems.