Random Neural Networks

elm_fig_rawLimitation on the computational resources is an issue faced in machine learning.  In response, random neural networks and non-linear projections can be used, e.g. to train single-layer neural networks very fast (often called extreme learning machines in recent literature) and yet to obtain satisfactory prediction performances.  I analysed their behaviour when the number of neurons becomes very large.  In particular, I shown that such models can be formulated in terms of a kernel for which an analytical form exists.


Related publications