home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!zaphod.mps.ohio-state.edu!cs.utexas.edu!ut-emx!utig.ig.utexas.edu!mrinal
- From: mrinal@utig.ig.utexas.edu (Mrinal Sen)
- Newsgroups: comp.ai.neural-nets
- Subject: Mean field theory, Hopfield NN etc.
- Message-ID: <86464@ut-emx.uucp>
- Date: 12 Jan 93 21:30:02 GMT
- Sender: news@ut-emx.uucp
- Reply-To: mrinal@utig.ig.utexas.edu (Mrinal Sen)
- Organization: Institute for Geophysics,UTexas Austin
- Lines: 26
-
-
-
- In a series of papers Peterson and his colleagues mapped optimization problem
- into Hopfield NN and derived mean field theory equations for simulated
- annealing (Peterson and Soderberg 1989, international journal of neural
- systems, 1, 3-22.). They have described applications to travelling
- salesman (TSP) and graph partioning problems. However, I have some diffculty
- in mapping a general optimization problem into Hopfield network. I will
- appreciate if someone could give me some tips on that.
-
- Let us consider the problem of finding global minimum of a multimodal
- multidimensional function, say f(x,y,z) where x, y, and z can take discrete
- values within prespecified range. I am finding it difficult to map it onto
- a Hopfield network i.e to identify the connectivity Tij. It seems that
- mapping onto Hopfield network is not a precondition to use the MFT equations.
- However, how then should the energy function be defined ?
-
- Please mail your reply preferably to
-
- mrinal@bullen.ig.utexas.edu
-
- - mrinal sen
- Institute for Geophysics
- University of Texas at Austin
- Austin, Tx 78759
- (512)834-2782
-