def evaluation(trainer):
num_sample = 10000
xp = gen.xp
#xs = []
"""
for i in range(0, num_sample, batchsize):
z = Variable(xp.asarray(gen.make_hidden(batchsize)))
with chainer.using_config('train', False), chainer.using_config('enable_backprop', False):
x = gen(z)
x = chainer.cuda.to_cpu(x.data)
xs.append(x)
xs = np.asarray(xs)
"""
z = Variable(xp.asarray(gen.make_hidden(num_sample)))
with chainer.using_config('train', False), chainer.using_config('enable_backprop', False):
x = gen(z)
xs = chainer.cuda.to_cpu(x.data)
real_data = GmmDataset(num_sample, 123, num_cluster=8, std=0.02, scale=2)._data
a, b = np.ones((num_sample,)) / num_sample, np.ones((num_sample,)) / num_sample
#print(xs)
#print(real_data)
M = ot.dist(xs, real_data)
M /= M.max()
distance = ot.emd2(a, b, M)
del xs
gc.collect()
del real_data
gc.collect()
#print(distance)
chainer.reporter.report({
'EMD': distance,
})
def evaluation(trainer):
num_sample = 10000
xp = gen.xp
#xs = []
"""
for i in range(0, num_sample, batchsize):
z = Variable(xp.asarray(gen.make_hidden(batchsize)))
with chainer.using_config('train', False), chainer.using_config('enable_backprop', False):
x = gen(z)
x = chainer.cuda.to_cpu(x.data)
xs.append(x)
xs = np.asarray(xs)
"""
z = Variable(xp.asarray(gen.make_hidden(num_sample)))
with chainer.using_config('train', False), chainer.using_config(
'enable_backprop', False):
x = gen(z)
xs = chainer.cuda.to_cpu(x.data)
real_data = GmmDataset(num_sample,
123,
num_cluster=8,
std=0.02,
scale=2)._data
a, b = np.ones((num_sample, )) / num_sample, np.ones(
(num_sample, )) / num_sample
#print(xs)
#print(real_data)
M = ot.dist(xs, real_data)
M /= M.max()
distance = ot.emd2(a, b, M)
del xs
gc.collect()
del real_data
gc.collect()
#path="/Users/keiikegami/Dropbox/research/Imaizumi-Sensei/paper_result/2dim/emd.pickle"
with open("emd2.txt", 'ab') as f:
f.write("{%d: %f}\n".encode() % (combination, distance))
def main():
parser = argparse.ArgumentParser(description='WGAN')
parser.add_argument('--batchsize', '-b', type=int, default=64,
help='Number of images in each mini-batch')
parser.add_argument('--datasize', '-d', type=int, default=10000,
help='Number of samples')
parser.add_argument('--lam', '-l', type=int, default=0.5,
help='Hyperparameter of gp')
parser.add_argument('--lam2', '-l2', type=int, default=1.0,
help='Hyperparameter of gp')
parser.add_argument('--n_hidden', type=int, default=100,
help='latent variable')
parser.add_argument('--seed', '-s', type=int, default=111,
help='seed number')
parser.add_argument('--epoch', '-e', type=int, default=300,
help='Number of sweeps over the dataset to train')
parser.add_argument('--gpu', '-g', type=int, default=-1,
help='GPU ID (negative value indicates CPU)')
parser.add_argument('--out', '-o', default='result',
help='Directory to output the result')
parser.add_argument('--resume', '-r', default='',
help='Resume the training from snapshot')
parser.add_argument('--unit', '-u', type=int, default=1000,
help='Number of units')
parser.add_argument('--setting', '-set', type=int, default=10,
help='Number of units')
args = parser.parse_args(args=[])
# parameter set
## Optimizers
alpha = 0.002
beta1 = 0.9
beta2 = 0.999
## Network
dim = 2
num_nets = 1
wscale = 0.02
# Networks
generator = Generator(dim, num_nets, args.n_hidden, wscale)
critic = Critic(num_nets, wscale)
if args.gpu >= 0:
chainer.cuda.get_device(args.gpu).use()
generator.to_gpu()
critic.to_gpu()
# Optimizer set
opt_g = chainer.optimizers.Adam(alpha = alpha, beta1 = beta1, beta2 = beta2, weight_decay_rate=0.0001)
opt_g.setup(generator)
opt_g.add_hook(chainer.optimizer.GradientClipping(1))
opt_c = chainer.optimizers.Adam(alpha = alpha, beta1 = beta1, beta2 = beta2, weight_decay_rate=0.0001)
opt_c.setup(critic)
opt_c.add_hook(chainer.optimizer.GradientClipping(1))
# Dataset
dataset = GmmDataset(args.datasize, args.seed, std=0.02, scale=2)
train_iter = chainer.iterators.SerialIterator(dataset, args.batchsize)
# Trainer
updater = WGANUpdater(train_iter, generator, critic, 5, opt_g, opt_c, args.lam, args.lam2, args.n_hidden, device=args.gpu)
trainer = training.Trainer(updater, (args.epoch, 'epoch'), out=args.out)
# Extensions
trainer.extend(extensions.LogReport())
trainer.extend(
extensions.PlotReport(['loss/critic'],
'epoch', file_name='criticloss_2dim%d.png' % args.setting))
trainer.extend(
extensions.PlotReport(
['loss/generator'], 'epoch', file_name='genloss_2dim%d.png' % args.setting))
trainer.extend(extensions.PrintReport(
['epoch', 'loss/critic', 'loss/generator', 'elapsed_time']))
out = "/Users/keiikegami/Dropbox/research/Imaizumi-Sensei/paper_result/2dim/result%d" % args.setting
trainer.extend(calcEMD(generator), trigger=(1, 'epoch'))
trainer.extend(extensions.LogReport(log_name = "log_2dim%d" % args.setting))
trainer.extend(out_generated(generator, args.seed + 10, out, radius=2), trigger=(10, 'epoch'))
trainer.extend(extensions.ProgressBar())
if args.resume:
chainer.serializers.load_npz(args.resume, trainer)
# Run
trainer.run()
fig, axes = plt.subplots(1, 1, figsize=(5, 5))
axes.scatter(data[:, 0], data[:, 1],
alpha=0.5, color='darkgreen', s=17)
axes.set_title('Real Data')
axes.set_xlim(-radius-5.0, radius+5.0)
axes.set_ylim(-radius-5.0, radius+5.0)
fig.tight_layout()
fig.savefig(file_path + '/training_data_scatter.jpg')
plt.close(fig)
if __name__ == "__main__":
gmm = GmmDataset(10000, 123, num_cluster=8, std=0.02, scale=2)
file = "/Users/keiikegami/Dropbox/research/Imaizumi-Sensei/image_results/2dim"
data = gmm._data
plot_kde_data_real(data, file, radius = 2)
plot_scatter_real_data(data, file, radius = 2)
def calcEMD(gen):
@chainer.training.make_extension()
def evaluation(trainer):
num_sample = 10000
xp = gen.xp
#xs = []
"""
for i in range(0, num_sample, batchsize):