At a glance... |
Syllabus |
Models |
Code |
Lecturer
After doing all the following, you should
be able to write one source files into hw/code/3 along with
screen snaps of your work (if relevant).
Using some URL shortener (e.g. goo.gl), shorten the URL to hw/code/4
and paste into the submission page.
Read the lecture on simulated annealing.
Hack up a simulated annealer for the Schaffer model.
Do not be clever. This is throw away code. As quick and as dirty as you like.
To create neighbors, just mutate the single independent variable at each run.
To compute energy, add the sum of the two f1, f2 variables. Note that for the lecture code to work, that sum has to be normalized 0..1. So conduct a baseline study where you run the Schaffer 100 times to find the min and max values for f1 + f2 . Take those numbers then hardwire the following normalization function:
energy(f1,f2) = ((f1 + f2) - min) / (max - min)
(Note also, you will need that maximum energy figure to implement emax).
Try and make a report that looks like the follow. Do not sweat the details (near enough is good enough but the idea is that every 25 evaluations, we print a new line and the current energy)/ Also, the idea here is that every try to make the crazy jumps ("?") less frequent later on (and don't worry if you cannot).
### saDemo ##################################################
# 2015-08-09 00:23:35
# Basic study.
!!! Schaffer
Schaffer
, 0000, :0.2, !.?+?+??.+?.++?+.++...?.+
, 0025, :1.0, .?+..?+?+..+..?++..??+.?.
, 0050, :1.0, +....?+.?+.+.?++++?+.?.+.
, 0075, :1.0, .?+......?+..?+......+...
, 0100, :1.0, .+........?+.?++++?.+?+..
, 0125, :1.0, ..+...?+.................
, 0150, :1.0, ................?++...+?.
, 0175, :1.0, +++...?+.+.?++.......?.+.
, 0200, :1.0, ...?+..+.+..?+.+.+.......
, 0225, :1.0, ...?+..............+...?+
, 0250, :1.0, .....+...+....?+.+..+....
, 0275, :1.0, ........................?
, 0300, :1.0, .++.+.............?+.+..+
, 0325, :1.0, .........................
, 0350, :1.0, .........................
, 0375, :1.0, .........................
, 0400, :1.0, ?++++............+.......
, 0425, :1.0, .........................
, 0450, :1.0, .........................
, 0475, :1.0, ........?+.?+............
, 0500, :1.0, ....?+.++.....+..........
, 0525, :1.0, ..........?.+............
, 0550, :1.0, .....................?++.
, 0575, :1.0, .++......................
, 0600, :1.0, .........................
, 0625, :1.0, .........................
, 0650, :1.0, ...........?+++.?+.+.+...
, 0675, :1.0, .........................
, 0700, :1.0, .....?+..................
, 0725, :1.0, .........................
, 0750, :1.0, .............?+..+.......
, 0775, :1.0, .........................
, 0800, :1.0, .........................
, 0825, :1.0, .........................
, 0850, :1.0, .........................
, 0875, :1.0, ...................?.+...
, 0900, :1.0, .....?+++...?++..+....+..
, 0925, :1.0, ?++.+.+..................
, 0950, :1.0, ....?.+...+.....+........
, 0975, :1.0, .........................
------
:e 0.997611905488
:y 0.490, 1.690
:x 0.700
------------------------------------------------------------------------
:cache
:keep 128 :pending 4
:misc
:a12 0.56 :copyleft SEARCH-BASED SE Tools
(c) 2014, copyright BSD-3, Tim Menzies :early 5 :epsilon 1.01 :era 25 :repeats 30 :seed 1 :verbose True
:sa
:baseline 100 :cooling 0.6 :kmax 1000 :patience 2000
# Runtime: 0.023 secs
```
## Tips
### Write, no new line
In olde Python, to print something without a new line:
```python
import sys
def say(x):
sys.stdout.write(str(x)); sys.stdout.flush()
```
In new Python:
```python
from __future__ import print
def say(x): print(x, end="")
```
### (Optional) Baselines
In the above, it was needed to pre-run the model a
hundred times or so to learn min,max and _emax_.
That is a simple enough process to automate.
_________
<img align=right src="https://raw.githubusercontent.com/txt/mase/master/img/pd-icon.png">Copyright © 2015 [Tim Menzies](http://menzies.us).
This is free and unencumbered software released into the public domain.
For more details, see the [license](https://github.com/txt/mase/blob/master/LICENSE.md).