itemsinXs = np.unique(
rw.flatten_list(Xs)) # list of items that appear in toy data
# reconstruct graph
rw_graph = rw.noHidden(Xs, numnodes)
ui_graph, bestval = rw.findBestGraph(Xs, numnodes=numnodes)
# remove nodes not in X from all graphs before comparison
rw_graph = rw_graph[itemsinXs, ]
rw_graph = rw_graph[:, itemsinXs]
ui_graph = ui_graph[itemsinXs, ]
ui_graph = ui_graph[:, itemsinXs]
usfcopy = np.copy(usf)
usfcopy = usfcopy[itemsinXs, ]
usfcopy = usfcopy[:, itemsinXs]
sdt_rw = rw.costSDT(rw_graph, usfcopy)
sdt_ui = rw.costSDT(ui_graph, usfcopy)
sdt_rws.append(sdt_rw)
sdt_uis.append(sdt_ui)
towrite = rw.flatten_list([sdt_rw, sdt_ui])
towrite = ','.join([str(i) for i in towrite])
towrite = str(numlists) + ',' + str(listlength) + ',' + str(
gnum) + ',' + towrite
f.write(towrite)
f.write('\n')
fe_graph = rw.firstEdge(flatdata, usf_numnodes)
if 'uinvite' in methods:
uinvite_graphs, priordict = rw.hierarchicalUinvite(
datab[:listnum],
items[:listnum],
numnodes[:listnum],
toydata,
fitinfo=fitinfo)
uinvite_group_graph = rw.priorToGraph(priordict,
usf_items,
cutoff=0.7) #JZ
for method in methods:
if method == "rw":
costlist = [
rw.costSDT(rw_graph, usf_graph),
rw.cost(rw_graph, usf_graph)
]
if method == "goni":
costlist = [
rw.costSDT(goni_graph, usf_graph),
rw.cost(goni_graph, usf_graph)
]
if method == "chan":
costlist = [
rw.costSDT(chan_graph, usf_graph),
rw.cost(chan_graph, usf_graph)
]
if method == "kenett":
costlist = [
rw.costSDT(kenett_graph, usf_graph),
graphs = alldata['uinvite_graphs']
items = alldata['items'][0:10]
priordict = alldata['priordict']
# for old pickles. ha, that's kinda funny.
if 'DEFAULTPRIOR' not in priordict.keys():
priordict['DEFAULTPRIOR'] = 0.5
# recompute for halfa... should be identical for a=1
#priordict = rw.genGraphPrior(graphs, items, a_inc=0.5)
print rw.probXhierarchical(Xs, graphs, items, priordict, td)
for cut in [j/20.0 for j in range(21)]:
uinvite_group_graph = rw.priorToGraph(priordict, usf_items,cutoff=cut)
asd=rw.costSDT(uinvite_group_graph, usf_graph)
print ",".join([str(i) for i in [filename.split('.')[0], cut, asd[0], asd[2], asd[1]+asd[2]]])
# newmixturepe3 >> 281
# newmixturep275 >> 275
# newmixturep4 + cutoff=.8 >> 267
# constructed with a_inc=1
# [109, 284, 33, 12294]
# -8531.6869171978178
# constructed with a_inc=0.5, prob checked with new prior a_inc=1.0
# [141, 252, 31, 12296]
# -8474.0454830370072
# constructed with a_inc=0.5
Xs = [[convertX[i] for i in x] for x in Xs]
datab.append(Xs)
seednum += numlists
listnum=10
uinvite_graphs, priordict = rw.hierarchicalUinvite(datab[:listnum], items[:listnum], numnodes[:listnum], toydata, fitinfo=fitinfo)
uinvite_group_graph = rw.priorToGraph(priordict, usf_items)
alldata=dict()
alldata['uinvite_graphs'] = uinvite_graphs
alldata['priordict'] = priordict
alldata['uinvite_group_graph'] = uinvite_group_graph
alldata['datab'] = datab
alldata['items'] = items
alldata['numnodes'] = numnodes
#alldata['td'] = toydata
#alldata['fitinfo'] = fitinfo
fh=open(filename,"w")
pickle.dump(alldata,fh)
fh.close()
costlist = [rw.costSDT(uinvite_group_graph, usf_graph), rw.cost(uinvite_group_graph, usf_graph)]
costlist = rw.flatten_list(costlist)
for i in costlist:
print i, ",",
print rw.probXhierarchical(datab[:listnum], uinvite_graphs[:listnum], items[:listnum], priordict, toydata)