LMAM/OLMAM Matlab Toolbox: The LMAM-OLMAM Matlab Toolbox is a set of Matlab routines that implement the Levenberg Marquardt with Adaptive Momentum (LMAM) and the Optimized Levenberg Marquardt with Adaptive Momentum (OLMAM) algorithms. These are two very efficient second order algorithms for training feedforward neural networks and, for some problems (Sonar data, 2-Spirals), they have been shown to achieve the best training results on standard benchmark datasets ever reported in the neural networks literature.
More details of the algorithms can be found in our paper appearing in IEEE Transactions on Neural Networks (September 2002):
N Ampazis and S. J. Perantonis: “Two Highly Efficient Second Order Algorithms for Training Feedforward Networks”, IEEE Transactions on Neural Networks Vol. 13, No. 5, September 2002.
You can download the toolbox from Github