-*- mode:Text -*-
Copyright 1992 Patrick H. Winston.  All rights reserved.
Version 1.1, transferred from master disk on 21 Jan 93
 
This file may reference other files containing copyrighted software.
Any restrictions on the use of such software are described in the
files containing that software.  This file may be freely copied as
long as this notice is kept intact.

PURPOSE

Provides program support for Chapter 19, Learning by Recording Cases.
Illustrates nearest neighbor learning using colored blocks.

PERFORMANCE EXPERIMENT 1

Examine the file kd.exp.  Run the nearest-neighbor identification
experiments by typing the following command to your lisp system:

(load "kd.exp")

Alternatively, evaluate each Lisp expression in the file one at a
time, noting what each does.

PERFORMANCE EXPERIMENT 2

Run the nearest-neighbor identification procedures with other
samples and/or with other unknowns.

MEDIUM-SCALE LISP PROJECT 1

Modify the program such that it uses samples to predict a number,
such as 10, instead of a symbol, such as BLUE.  When your
modified program is instructed to use more than one nearest
neighbor, have your program average the values provided by the
nearest neighbors.

MEDIUM-SCALE LISP PROJECT 2

Again, modify the program such that it uses samples to predict a
number, such as 10, instead of a symbol, such as BLUE.  This
time, however, when your modified program is instructed to use
more than one nearest neighbor, have your program fit a surface
to the predictions made by the nearest neighbors---using say,
least squares---and then have your program find the appropriate
point on that surface.