Sound source ranging using a feed-forward neural network with fitting-based early stopping
This work addresses a specific challenge in underwater acoustics for researchers, but it is incremental as it builds on existing neural network methods.
The paper tackled the problem of evaluating range accuracy for feed-forward neural networks in ocean waveguide source ranging by introducing a fitting-based early stopping method, which improved ranging accuracy on test data with unknown distances, as demonstrated on simulated and experimental data.
When a feed-forward neural network (FNN) is trained for source ranging in an ocean waveguide, it is difficult evaluating the range accuracy of the FNN on unlabeled test data. A fitting-based early stopping (FEAST) method is introduced to evaluate the range error of the FNN on test data where the distance of source is unknown. Based on FEAST, when the evaluated range error of the FNN reaches the minimum on test data, stopping training, which will help to improve the ranging accuracy of the FNN on the test data. The FEAST is demonstrated on simulated and experimental data.