In [1]:
import math
import random
import copy
import sys, os
import tree
import Queue
In [2]:
def evaluate(x, formula):
local = x
print "local is: ", local
print "formula is: ",formula
ret = eval(formula)
print "ret is ", ret
return ret
In [3]:
goodbad = [True, False]
alltestssofar = []
numtests = 10
initialexperiments = []
fileicareabout = open("summary.txt","r")
text = fileicareabout.readlines()
fileicareabout.close()
workflow = text[0]
if (workflow != "null\n"):
script,func = (workflow[:-1]).split(",")
workflow = getattr(__import__(script), func)
else:
workflow = None
formula = text[1]
cols = text[2][:-1].split(",")
alllines = text[3:]
allexperiments = []
allresults = [] # experiments and their results
believeddecisive = []
In [4]:
moreexperiments = []
for e in alllines:
exp = (e[:-1]).split(",")
moreexperiments.append(exp)
for i in range(len(moreexperiments) / 10):
exp = random.choice(moreexperiments)
allexperiments.append(exp)
initialexperiments.append(exp)
rets=[]
In [5]:
for e in initialexperiments:
x = copy.deepcopy(e)
x[-1] = evaluate(e, formula)
allresults.append(x)
print "allresults is: ", allresults
Now the let's use the decision tree¶
In [6]:
def determinenodepurity(node, path):
print "path is: ", path
if node.results is None:
pathfalse = copy.deepcopy(path)
pathfalse.append((node.col,node.value,False))
path.append((node.col,node.value,True))
determinenodepurity(node.fb,pathfalse)
determinenodepurity(node.tb,path)
elif (len(node.results.items())>1):
mixedlist.append(path)
elif(node.results.items()[0][0]):
puregoodlist.append(path)
else:
purebadlist.append(path)
def findshoertestpaths(node):
q = Queue.Queue()
q.put((node,[]))
puregoodpath = None
purebadpath = None
while (not q.empty()) and ((puregoodpath is None) or (purebadpath is None)):
current = q.get()
if current[0].results is None:
q.put((current[0].fb,current[1]+[(current[0].col,current[0].value,False)]))
q.put((current[0].tb,current[1]+[(current[0].col,current[0].value,True)]))
elif (len(current[0].results.items())>1):
continue
elif(current[0].results.items()[0][0]) and (puregoodpath is None):
puregoodpath = current[1]
elif (not current[0].results.items()[0][0]) and (purebadpath is None):
purebadpath = current[1]
return [puregoodpath, purebadpath]
In [7]:
def manufacturetests(moralflag, alist):
global believeddecisive
rebuild = False;
print "a list: ", alist
if (moralflag == 'bad'):
for path in alist:
print "at step 2.5, path: ", path
z = assembletests('bad', path)
if z:
print "believeddecisive for bad: ",path
else:
rebuild = True
if (moralflag == 'good'):
for path in alist:
print "at step 3.5, path: ", path
z = assembletests('good', path)
if z:
print "believeddecisive for good: ",path
else:
rebuild = True
return rebuild
def assembletests(moralflag, path):
global believeddecisive
outlist = []
myarrtrans = zip(*allexperiments)
for j in range(len(cols) - 1):
y = set(myarrtrans[j])
print "y is: ", y
for index,value,flag in [tup for tup in path if tup[0] == j]:
if flag:
y = {value}
else:
y = y - {value}
outlist.append(list(y))
print "in assembletests, outlist is: ", outlist
# Now just choose random values from these lists to generate experiments
experiments = []
for i in range(numtests):
x = []
for mylist in outlist:
x.append(random.choice(mylist))
length = len(x)
for e in moreexperiments:
if x == e[:-1]:
x = copy.deepcopy(e)
if (x not in allexperiments) and (x not in experiments) and (len(x) == length + 1):
print("retrieved experiment: ", x)
experiments.append(x)
elif (workflow) and (len(x) == length):
print("called workflow")
x = workflow(x)
experiments.append(x)
else:
print "experiment already seen: ", x
print "experiments are: ", experiments
# simulate this just for experiments
# then if you get the answer you expect every time, then modify believeddecisive
# try experiments in the simulator and evaluate
# FOLLOWING LOOP PMUST CHANGE WHEN WE GO AWAY FROM SIMULATION
# RESULTS MUST BE READ BACK INTO allrets AND allresults SHOULD BE APPENDED TO
# BECAUSE allresults HAS THE RESULTS OF ALL EXPERIMENTS
for e in experiments:
# use experiments as long as labels so far are unambiguous
allexperiments.append(e)
x = copy.deepcopy(e)
x[-1] = (evaluate(e, formula))
allresults.append(x)
if ((moralflag == 'good') and (x[-1] == goodbad[1])) or ((moralflag == 'bad') and (x[-1] == goodbad[0])) :
return False
return True
In [8]:
rebuild = True
while rebuild:
rebuild = False
print "allresults is: ", allresults
print "allexperiments are: ", allexperiments
t = tree.build(allresults, cols=cols)
tree.draw_tree(t)
#puregoodlist = []
#purebadlist = []
#mixedlist = []
goodpath, badpath = findshoertestpaths(t)
rebuild = rebuild or manufacturetests('good', [goodpath])
rebuild = rebuild or manufacturetests('bad', [badpath])
print len(allexperiments), "experiments so far"
In [ ]: