This tool trains a three-layered network of sigmoidal units using back-propagation to classify fish according to their lengths as shown below.

FIGURE 9.1 is a flow diagram illustrating a procedure for classification of fishes. In this case the fish species to be classified are salmon and trout. (Diagram after Duda et al. 2001; drawings courtesy of Bob Hines, U.S. Fish and Wildlife Service.)

FIGURE 9.2 models a three-layered, feedforward network used to classify fish by length There is 1 input unit x, 12 hidden units yi, and 3 output units z1, z2, and z3. The hidden, input, and output units are indexed by i (i= 1,.. .,12), j (j = 1), and k (k = 1,2,3), respectively. The input unit projects to the hidden units over the weights in matrix V (with elements vij). In this case, in which there is only one input unit, V is a column vector. The hidden units project to the output units over the weights in matrix U (with elements uki). The units are sigmoidal (see Chaper 6).

This tool is built from the MATLAB scripts for Tutorial on Neural Systems Modeling by Thomas J. Anastasio.

Anastasio, T. J. (2009). Tutorial on Neural Systems Modeling. Sinauer Associates, Incorporated.

Researchers should cite this work as follows:

• Lisa Sproat (2013), "[Illinois]: Fish classification using back-propagation," https://nanohub.org/resources/fishbackprop. (DOI: 10.4231/D32R3NX1X).

  BibTex | EndNote

1. nano/bio
2. neuroscience
3. teaching materials