I thought I’d share these papers since I put some time into them, I find them interesting, and there seems to be a general dearth of information on the internet regarding competitive learning networks and Hopfield networks. This is the class I took which was fairly interesting and informative though I was dissapointed that it didn’t cover reinforcement learning.
Clustering Animals and Congressmen through a Competitive Learning Network (PDF)
This paper explains the results and process of running an unsupervised learning technique
called a competitive learning network first on a data set of 10 animals and their various features. We wish
to see how well the competitive learning network segments the animals into their
respective biological classes (e.g. birds, mammals, and reptiles). This is a reproduction
of the experiment done in Knight . Next I look at how well the competitive learning
network performs on a data set composed of congressmen and how they voted on 16 bills
in 1984 . I am especially curious here to see if the competitive learning network can
cluster the congress men into their respective political parties (e.g. Democrat, and
Hopfield Networks: a Simple OCR Application (PDF)
In this paper I discuss the results of training and testing a Hopfield network and its optimized counterpart to recognize the digits 0-9. I also compare the performance of the optimized and non-optimized brands.
This paper covers my own experience of following the process laid out in Sigillito.
My results turn out to be very similar to those found by Sigilitto.
My professor provided the software used in these papers. He probably wouldn’t want me to offer the software for the world to download all willy-nilly on here but I’m sure providing it on a case by case basis is fine. So if you want to use it, just send me and email and I’ll send it to you.
Tags: neural networks, neural networks class, nueral networks, neural network, competitive learning network, competitive learning networks, CLN, Hopfield, Hopfield Net, Hopfield Nets, Hopfield Network, Hopfield Networks, Machine Learning