Ole Christian Eidheim

Ole Christian Eidheim

There is no convincing evidence that backpropagation is a biologically plausible mechanism, and further studies of alternative learning methods are needed. A novel online clustering algorithm is presented that can produce arbitrary shaped clusters from inputs in an unsupervised manner, and requires no prior knowledge of the number of clusters in the input data. This is achieved by finding correlated outputs from functions that capture commonly occurring input patterns. The algorithm can be deemed more biologically plausible than model optimization through backpropagation, although practical applicability may require additional research. However, the method yields satisfactory results on several toy datasets on a noteworthy range of hyperparameters.

Ole Christian Eidheim

Despite the recent success of artificial neural networks, more biologically plausible learning methods may be needed to resolve the weaknesses of backpropagation trained models such as catastrophic forgetting and adversarial attacks. Although these weaknesses are not specifically addressed, a novel local learning rule is presented that performs online clustering with an upper limit on the number of clusters to be found rather than a fixed cluster count. Instead of using orthogonal weight or output activation constraints, activation sparsity is achieved by mutual repulsion of lateral Gaussian neurons ensuring that multiple neuron centers cannot occupy the same location in the input domain. An update method is also presented for adjusting the widths of the Gaussian neurons in cases where the data samples can be represented by means and variances. The algorithms were applied on the MNIST and CIFAR-10 datasets to create filters capturing the input patterns of pixel patches of various sizes. The experimental results demonstrate stability in the learned parameters across a large number of training samples.

Dherik Devakumar, Ole Christian Eidheim

A novel wavelet-like function is presented that makes it convenient to create filter banks given mainly two parameters that influence the focus area and the filter count. This is accomplished by computing the inverse Fourier transform of Gaussian functions on logarithmic frequency axes in the frequency domain. The resulting filters are similar to Gabor filters and represent oriented brief signal oscillations of different sizes. The wavelet-like function can be thought of as a generalized Log-Gabor filter that is multidimensional, always uses Gaussian functions on logarithmic frequency axes, and innately includes low-pass filters from Gaussian functions located at the frequency domain origin.