CPSC 420-500 Read-Only Bulletin Board

[Q] How do I calculate "target-output"? What is target and what is output?

[A] target is the desired output value, and output is the output generated
by the perceptron, given the current connection weights.

For example, here's a training set (table below). The right most column shows 
the "target" values for the boolean function AND.

x       y       | AND
----------------+----
0       0          0
0       1          0
1       0          0
1       1          1

First row: input 0 0, target 0.  Last row: input is 1 1, the target is 1. 
Output is what you get from your perceptron when you plug in 0 0. It is the 
output of the perceptron for the current set of weights. 

Most of the papers you should be able to find on the web without a problem.
Here are the main ones.

Bell 1999

http://www.cnl.salk.edu/~tony/ptrsl.pdf

Dean 1998

http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6VH9-3SX87GK-5&_us
er=952835&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000049198&_versio
n=1&_urlVersion=0&_userid=952835&md5=7d848a80f59cc504f37d8a1e1df42160

* you need to be on the campus network to download the link above.

Cohen and Beal 2000

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.32.7512&rep=rep1&ty
pe=pdf

Choe and Smith 2006

http://faculty.cs.tamu.edu/choe/ftp/publications/choe.aaai06.pdf

Beer 2000

http://eecs.case.edu/~beer/Papers/TICS.pdf

Kwon and Choe 2008

http://faculty.cs.tamu.edu/choe/ftp/publications/kwon.icdl08.pdf

Here's how you compile:

; compile
(compile-file "dfs.lsp")

; load the compiled binary
(load "dfs.sparcf")

; run the function
(dfs '((2 9) (4 (1 30) 2) 5))


Here's how you compile, with tail-recursion turned off:

; compile with tail-recursion off
(compile-file "dfs.lsp" :block-compile nil)

; load the compiled binary
(load "dfs.sparcf")

; turn trace on on the recusrive function
(trace dfs-core)

; run the function
(dfs '((2 9) (4 (1 30) 2) 5))



Here's an example run. Bold is what you enter in the Lisp interpreter.

* (compile-file "dfs.lsp" :block-compile nil)

; Python version 1.1, VM version UltraSparc/Solaris 7 on 16 SEP 08 09:51:43 pm.
; Compiling: /home/faculty/choe/web_project/625/src/dfs.lsp 25 AUG 08 08:57:56 pm

; Converted DFS.
; Compiling DEFUN DFS:
; Converted DFS-CORE.
; Compiling DEFUN DFS-CORE:
; Converted GOALP.
; Compiling DEFUN GOALP:
; Converted MAKE-NODE-LIST.
; Compiling DEFUN MAKE-NODE-LIST:
; Converted EXPAND.
; Compiling DEFUN EXPAND:
; Byte Compiling Top-Level Form:

; dfs.sparcf written.
; Compilation finished in 0:00:00.

#P"/home/faculty/choe/web_project/625/src/dfs.sparcf"
NIL
NIL
* (load "dfs.sparcf")

; Loading #P"/home/faculty/choe/web_project/625/src/dfs.sparcf".
T
* (dfs '((2 9) (4 (1 30) 2) 5))

  0: (DFS-CORE ((# # 5)))
  0: DFS-CORE returned 30
30
* (trace dfs-core)
Warning:  Function DFS-CORE already TRACE'd, retracing it.

(DFS-CORE)
*  (dfs '((2 9) (4 (1 30) 2) 5))

  0: (DFS-CORE ((# # 5)))
    1: (DFS-CORE ((2 9) (4 # 2) 5))
      2: (DFS-CORE (2 9 (4 # 2) 5))
        3: (DFS-CORE (9 (4 # 2) 5))
          4: (DFS-CORE ((4 # 2) 5))
            5: (DFS-CORE (4 (1 30) 2 5))
              6: (DFS-CORE ((1 30) 2 5))
                7: (DFS-CORE (1 30 2 5))
                  8: (DFS-CORE (30 2 5))
                  8: DFS-CORE returned 30
                7: DFS-CORE returned 30
              6: DFS-CORE returned 30
            5: DFS-CORE returned 30
          4: DFS-CORE returned 30
        3: DFS-CORE returned 30
      2: DFS-CORE returned 30
    1: DFS-CORE returned 30
  0: DFS-CORE returned 30
30
* 

1. [Q] How do I deal with division by zero?

   [A] You don't need to be too thorough on this. Just check for obvious cases in sdiv. For example, if you got (/ x (- 1 1)) as a result of differentiation, trying to evaluate it will give you a divide-by-zero error. This cannot be checked easily in sdiv. So, don't worry about such a case. You need to take care of (/ x 0) though.

2. [Q] How far should simplification go?

   [A] Just use the ones mentioned in prog1.pdf, those involving identities and/or numbers. You don't need to consider things like x + x to 2x.

3. [Q] How do I do deriv-eval?

   [A] There are two ways, a simple approach using (eval (list ....)), and a more complex one using recursive expression evaluation. Either one is fine. When doing deriv-eval, assume that only numbers and the differentiation variable are used. For example, avoid something like (+ x (* y 3)) since you'll get an error trying to evaluate y.

Nothing here yet.