#Es. create non overlapping set
s = data3.shuffle()
s1 = data3[0:50]
s2 = data3[50:100]
s3 = data3[100:]
rawDat_s1,rawLabels_s1 = [s1.data,s1.labels]
rawDat_s2,rawLabels_s2 = [s2.data,s2.labels]
#GRAPH
#download from TUDataset repository
#TODO: will be change
datasets.get_dataset("COIL-RAG")
gdata1 = graph_nxDataset("/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/COIL-RAG", "Letter-h", readergraph, pre_transform=gTransfItem())
gdata2 = graph_nxDataset("/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/COIL-RAG", "Letter-h", readergraph, transform = gTransfItem())
gdata3 = graph_nxDataset("/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/COIL-RAG", "Letter-h", readergraph)
print("original data get item: ", gdata3[0].x.nodes[0]['attributes'])
print("transform data get item: ", gdata2[0].x.nodes[0]['attributes'])
print("pre_transformed data get item:", gdata1[0].x.nodes[0]['attributes'])
print("original data stored value: ", gdata3.data[0].nodes[0]['attributes'])
print("transform data stored value: ", gdata2.data[0].nodes[0]['attributes'])
print("pre_transformed data stored value: ", gdata1.data[0].nodes[0]['attributes'])
# ####
# data1_perm = data1.shuffle()
# data0_key = data1_perm.indices[0]
def edgeDissimilarity(a, b):
D = 0
if (a['labels'] != b['labels']):
D = 1
return D
def readergraph(path):
graphs_nx = reader.tud_to_networkx("Mutagenicity")
classes = [g.graph['classes'] for g in graphs_nx]
return graphs_nx, classes
print("Loading...")
data1 = graph_nxDataset(
"/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/Mutagenicity",
"Mutagenicity", readergraph)
#not connected graph
cleanData = [(g, idx, label)
for g, idx, label in zip(data1.data, data1.indices, data1.labels)
if nx.is_connected(g)]
cleanData = np.asarray(cleanData, dtype=object)
data1 = graph_nxDataset([cleanData[:, 0], cleanData[:, 2]], "Mutagenicity")
data1 = data1[0:10]
graphDist = BMF(nodeDissimilarity, edgeDissimilarity)
x1 = graphDist.pdist(data1.data, forceSym=True)
x2 = graphDist.pdist(data1.data, forceSym=False)
# @jitclass
def nodeDissimilarity(a, b):
return np.linalg.norm(np.asarray(a['attributes']) - np.asarray(b['attributes'])) / np.sqrt(2)
# @jitclass
def edgeDissimilarity(a,b):
return 0.0
def readergraph(path):
graphs_nx = reader.tud_to_networkx("Letter-high")
classes = [g.graph['classes'] for g in graphs_nx]
return graphs_nx, classes
print("Loading...")
data1 = graph_nxDataset("/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/Letter-high", "LetterH", reader = readergraph)
#Removed not connected graph and null graph!
cleanData=[]
for g,idx,label in zip(data1.data,data1.indices,data1.labels):
if not nx.is_empty(g):
if nx.is_connected(g):
cleanData.append((g,idx,label))
#cleanData = [(g,idx,label) for g,idx,label in zip(data1.data,data1.indices,data1.labels) if not nx.is_empty(g)]
cleanData = np.asarray(cleanData,dtype=object)
data1 = graph_nxDataset([cleanData[:,0],cleanData[:,2]],"Letter")
data1= data1[0:100]
#Test extr indices
data1 = data1.shuffle()
def main():
# def eaSimple(population, toolbox, cxpb, mutpb, ngen, stats=None,
# halloffame=None, verbose=__debug__):
####################
random.seed(64)
verbose = True
population = toolbox.population(n=POP_SIZE)
cxpb = CXPROB
mutpb = MUTPROB
ngen = N_GEN
halloffame = tools.HallOfFame(1)
stats = tools.Statistics(lambda ind: ind.fitness.values)
stats.register("avg", np.mean)
stats.register("std", np.std)
stats.register("min", np.min)
stats.register("max", np.max)
logbook = tools.Logbook()
logbook.header = ['gen', 'nevals'] + (stats.fields if stats else [])
###################
print("Loading...")
data1 = graph_nxDataset(
"/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/tudataset/Mutagenicity",
"Mutagenicity", readergraph)
#not connected graph
cleanData = [
(g, idx, label)
for g, idx, label in zip(data1.data, data1.indices, data1.labels)
if nx.is_connected(g)
]
cleanData = np.asarray(cleanData, dtype=object)
data1 = graph_nxDataset([cleanData[:, 0], cleanData[:, 2]], "Mutagenicity")
extract_func = randomwalk_restart.extr_strategy(max_order=6)
subgraph_extr = Extractor(extract_func)
# Evaluate the individuals with an invalid fitness
print("Initializing population...")
subgraphs = [
subgraph_extr.randomExtractDataset(data1, 1000) for _ in population
]
# mapWithConst = partial(toolbox.evaluate,granulationBucket=subgraphs)
invalid_ind = [ind for ind in population if not ind.fitness.valid]
fitnesses = toolbox.map(toolbox.evaluate, zip(invalid_ind, subgraphs))
for ind, fit in zip(invalid_ind, fitnesses):
ind.fitness.values = fit
if halloffame is not None:
halloffame.update(population)
record = stats.compile(population) if stats else {}
logbook.record(gen=0, nevals=len(invalid_ind), **record)
if verbose:
print(logbook.stream)
# Begin the generational process
print("Start evolution")
for gen in range(1, ngen + 1):
# Select the next generation individuals
offspring = toolbox.select(population, len(population))
# Vary the pool of individuals
offspring = varAnd(offspring, toolbox, cxpb, mutpb)
# Evaluate the individuals with an invalid fitness
subgraphs = [
subgraph_extr.randomExtractDataset(data1, 1000) for _ in population
]
invalid_ind = [ind for ind in offspring if not ind.fitness.valid]
#mapWithConst = partial(toolbox.evaluate,granulationBucket=subgraphs)
fitnesses = toolbox.map(toolbox.evaluate, zip(invalid_ind, subgraphs))
for ind, fit in zip(invalid_ind, fitnesses):
ind.fitness.values = fit
# Update the hall of fame with the generated individuals
if halloffame is not None:
halloffame.update(offspring)
# Replace the current population by the offspring
population[:] = offspring
# Append the current generation statistics to the logbook
record = stats.compile(population) if stats else {}
logbook.record(gen=gen, nevals=len(invalid_ind), **record)
if verbose:
print(logbook.stream)
return population, logbook
def main():
# """
# Reproducing GRALG optimization with mu+lambda strategy of evolution
# """
####################
random.seed(64)
verbose = True
population = toolbox.population(n=MU)
cxpb = CXPROB
mutpb = MUTPROB
ngen = N_GEN
mu = MU
lambda_ = LAMBDA
halloffame = tools.HallOfFame(1)
stats = tools.Statistics(lambda ind: ind.fitness.values)
stats.register("avg", np.mean)
stats.register("std", np.std)
stats.register("min", np.min)
stats.register("max", np.max)
logbook = tools.Logbook()
logbook.header = ['gen', 'nevals'] + (stats.fields if stats else [])
###################
print("Loading...")
data1 = graph_nxDataset(
"/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/IAM/Letter3/Training/",
"LetterH",
reader=IAMreadergraph)
data2 = graph_nxDataset(
"/home/luca/Documenti/Progetti/E-ABC_v2/eabc_v2/Datasets/IAM/Letter3/Validation/",
"LetterH",
reader=IAMreadergraph)
# data1 = data1.shuffle()
# data2 = data2.shuffle()
#Removed not connected graph and null graph!
cleanData = []
for dataset in [data1, data2]:
for g, idx, label in zip(dataset.data, dataset.indices,
dataset.labels):
if not nx.is_empty(g):
if nx.is_connected(g):
cleanData.append((g, idx, label))
cleanData = np.asarray(cleanData, dtype=object)
normalize('coords', cleanData[:750, 0], cleanData[750:, 0])
#Slightly different from dataset used in pygralg
dataTR = graph_nxDataset([cleanData[:750, 0], cleanData[:750, 2]],
"LetterH",
idx=cleanData[:750, 1])
dataVS = graph_nxDataset([cleanData[750:, 0], cleanData[750:, 2]],
"LetterH",
idx=cleanData[750:, 1])
del data1
del cleanData
print("Setup...")
extract_func = randomwalk_restart.extr_strategy(max_order=6)
# extract_func = breadthFirstSearch.extr_strategy(max_order=6)
# subgraph_extr = Extractor(extract_func)
subgraph_extr = Extractor(extract_func)
expTRSet = dataTR.fresh_dpcopy()
for i, x in enumerate(dataTR):
k = 0
while (k < 50):
for j in range(1, 6):
subgraph_extr.max_order = j
expTRSet.add_keyVal(dataTR.to_key(i), subgraph_extr.extract(x))
k += 6
expVSSet = dataVS.fresh_dpcopy()
for i, x in enumerate(dataVS):
k = 0
while (k < 50):
for j in range(1, 6):
subgraph_extr.max_order = j
expVSSet.add_keyVal(dataVS.to_key(i), subgraph_extr.extract(x))
k += 6
# Evaluate the individuals with an invalid fitness
print("Initializing population...")
subgraphs = subgraph_extr.randomExtractDataset(dataTR, 1260)
invalid_ind = [ind for ind in population if not ind.fitness.valid]
fitnesses = toolbox.map(
functools.partial(toolbox.evaluate,
granulationBucket=subgraphs,
trEmbeddBucket=expTRSet,
vsEmbeddBucket=expVSSet,
TRindices=dataTR.indices,
VSindices=dataVS.indices,
TRlabels=dataTR.labels,
VSlabels=dataVS.labels), invalid_ind)
for ind, fit in zip(invalid_ind, fitnesses):
ind.fitness.values = fit
if halloffame is not None:
halloffame.update(population)
record = stats.compile(population) if stats else {}
logbook.record(gen=0, nevals=len(invalid_ind), **record)
if verbose:
print(logbook.stream)
# Begin the generational process
for gen in range(1, ngen + 1):
# Vary the population
offspring = varOr(population, toolbox, lambda_, cxpb, mutpb)
#Evaluate invalid of modified individual
invalid_ind = [ind for ind in offspring if not ind.fitness.valid]
fitnesses = toolbox.map(
functools.partial(toolbox.evaluate,
granulationBucket=subgraphs,
trEmbeddBucket=expTRSet,
vsEmbeddBucket=expVSSet,
TRindices=dataTR.indices,
VSindices=dataVS.indices,
TRlabels=dataTR.labels,
VSlabels=dataVS.labels), invalid_ind)
for ind, fit in zip(invalid_ind, fitnesses):
ind.fitness.values = fit
# Update the hall of fame with the generated individuals
if halloffame is not None:
halloffame.update(offspring)
# Select the next generation population
population[:] = toolbox.select(population + offspring, mu)
# Update the statistics with the new population
record = stats.compile(population) if stats is not None else {}
logbook.record(gen=gen, nevals=len(invalid_ind), **record)
if verbose:
print(logbook.stream)
return population, logbook