def generate_posterior(self,
num_dimensions=20,
known_active_dimensions=3,
num_data=100,
labels=None,
num_samples=1000,
num_warmup=500,
num_chains=1,
device='gpu'):
X = self.X
Y = self.Y
hypers = self.hypers
sigma = hypers['sigma']
# args -- needed: num-chains
# num_dimensions
# active dimensions
if labels is None:
labs = [str(_) for _ in range(X.shape[1])]
else:
labs = labels
# TODO:set up Numpyro device (should maybe do this in main script ?)
numpyro.set_platform(device)
numpyro.set_host_device_count(num_chains)
# do inference
rng_key = random.PRNGKey(self.seed)
samples = self.run_inference(self.model,
rng_key,
X,
Y,
hypers,
num_warmup=num_warmup,
num_chains=num_chains,
num_samples=num_samples)
# compute the mean and square root variance of each coefficient theta_i
means, stds = vmap(
lambda dim: self.analyze_dimension(samples, X, Y, dim, hypers))(
np.arange(X.shape[1]))
num_dims = len(means)
active_dimensions = []
for dim, (mean, std) in enumerate(zip(means, stds)):
# we mark the dimension as inactive if the interval [mean - 3 * std, mean + 3 * std] contains zero
lower, upper = mean - sigma * std, mean + sigma * std
inactive = "inactive" if lower < 0.0 and upper > 0.0 else "active"
if inactive == "active":
active_dimensions.append(dim)
print("[dimension %02d/%02d] %s:\t%.2e +- %.2e" %
(dim + 1, X.shape[1], inactive, mean, std))
print("Identified a total of %d active dimensions." %
(len(active_dimensions)))
# Compute the mean and square root variance of coefficients theta_ij for i,j active dimensions.
# Note that the resulting numbers are only meaningful for i != j.
if len(active_dimensions) > 0:
dim_pairs = np.array(
list(itertools.product(active_dimensions, active_dimensions)))
means, stds = vmap(
lambda dim_pair: self.analyze_pair_of_dimensions(
samples, X, Y, dim_pair[0], dim_pair[1], hypers))(
dim_pairs)
# print(dim_pairs)
dim_pair_arr = []
dim_pair_index = num_dims - 1
dim_pair_name = []
pair_labs = []
for dim_pair, mean, std in zip(dim_pairs, means, stds):
dim1, dim2 = dim_pair
if dim1 >= dim2:
continue
dim_pair_index += 1
lower, upper = mean - sigma * std, mean + sigma * std
if not (lower < 0.0 and upper > 0.0):
dim_pair_arr.append(dim_pair_index)
dim_pair_name.append('%d and %d' % (dim1 + 1, dim2 + 1))
format_str = "Identified pairwise interaction between dimensions %d and %d: %.2e +- %.2e"
print(format_str % (dim1 + 1, dim2 + 1, mean, std))
pair_labs.append(str(labs[dim1]) + ' + ' + str(labs[dim2]))
# Draw a single sample of coefficients theta from the posterior, where we return all singleton
# coefficients theta_i and pairwise coefficients theta_ij for i, j active dimensions. We use the
# final MCMC sample obtained from the HMC sampler.
## Get posterior samples from the sample_theta_space_modified() method
thetas = self.sample_theta_posterior(
X, Y, active_dimensions, samples['msq'][-1],
samples['lambda'][-1], samples['eta1'][-1],
samples['xisq'][-1], hypers['c'], samples['var_obs'][-1],
num_samples, dim_pair_arr)
print("Active dimensions: " + str(active_dimensions))
## Visualize the posterior from the example with corner
labels = ['dim ' + str(i) for i in active_dimensions]
active_dimensions = active_dimensions + dim_pair_arr
if len(dim_pair_name) != 0:
for n in range(len(dim_pair_name)):
labels.append('dim ' + dim_pair_name[n])
#fig = corner.corner(thetas, labels = labels);
return active_dimensions, thetas, labels, pair_labs
else:
return active_dimensions, [], []
if __name__ == "__main__":
assert numpyro.__version__.startswith("0.8.0")
parser = argparse.ArgumentParser(
description="Bayesian Models of Annotation")
parser.add_argument("-n",
"--num-samples",
nargs="?",
default=1000,
type=int)
parser.add_argument("--num-warmup", nargs="?", default=1000, type=int)
parser.add_argument("--num-chains", nargs="?", default=1, type=int)
parser.add_argument(
"--model",
nargs="?",
default="ds",
help='one of "mn" (multinomial), "ds" (dawid_skene), "mace",'
' "hds" (hierarchical_dawid_skene),'
' "id" (item_difficulty), "lre" (logistic_random_effects)',
)
parser.add_argument("--device",
default="cpu",
type=str,
help='use "cpu" or "gpu".')
args = parser.parse_args()
numpyro.set_platform(args.device)
numpyro.set_host_device_count(args.num_chains)
main(args)
randwalk = numpyro.sample('mu', dist.Normal(loc=0, scale=sd_randwalk))
value = X + randwalk
sd_value = numpyro.sample('sd', dist.Uniform(low=0.0, high=100.0))
pred_y = numpyro.sample('pred_y',
dist.Normal(loc=value, scale=sd_value),
obs=y)
if __name__ == "__main__":
# 初期設定
num_samples = 1000
num_warmup = 100
num_chains = 1
device = 'cpu'
rng_key = random.PRNGKey(0)
numpyro.set_platform(device)
numpyro.set_host_device_count(num_chains)
# ランダムウォーク波形生成
random_walk = jnp.cumsum(0.5 * np.random.randn(100000))
# 価格がランダムな場合を想定
# random_walk = 0.5*np.random.randn(100000)
random_walks = {
1: random_walk,
2: random_walk[::2],
4: random_walk[::2][::2],
8: random_walk[::2][::2][::2],
16: random_walk[::2][::2][::2][::2],
32: random_walk[::2][::2][::2][::2][::2],
def main(params):
#X, Y, expected_thetas, expected_pairwise = get_data(N=args.num_data, P=args.num_dimensions,
# S=args.active_dimensions)
fxdata5 = pd.read_csv('fxdata5.csv', header=0)
df = fxdata5.copy()
df = df.iloc[-64:,:]
drop_cols = list(df.columns[df.columns.str.contains('Dummy')])
drop_cols = drop_cols + list(df.columns[df.columns.str.contains('LogReturn')])
drop_cols = drop_cols + list(df.columns[df.columns.str.contains('Spot')])
# drop_cols = drop_cols + list(df.columns[df.columns.str.contains('HAR')])
df.drop(drop_cols, axis=1, inplace=True)
target_col = 'EURGBP_log_RealVol'
df['Target_' + target_col] = df[target_col].shift(1).copy()
cols = df.columns.tolist()
cols.insert(0, cols.pop(cols.index('Target_' + target_col)))
df = df[cols].set_index('Date').dropna()
df[df.columns] = StandardScaler().fit_transform(df) # scale
Y = np.array(df.iloc[:,0])
X = np.array(df.iloc[:,1:])
parser = argparse.ArgumentParser(description="Gaussian Process example")
parser.add_argument("-n", "--num-samples", nargs="?", default=8000, type=int)
parser.add_argument("--num-warmup", nargs='?', default=500, type=int)
parser.add_argument("--num-chains", nargs='?', default=1, type=int)
parser.add_argument("--num-data", nargs='?', default=len(X), type=int)
parser.add_argument("--num-dimensions", nargs='?', default=len(X.T), type=int)
parser.add_argument("--active-dimensions", nargs='?', default=3, type=int)
parser.add_argument("--device", default='cpu', type=str, help='use "cpu" or "gpu".')
args = parser.parse_args(args=[])
numpyro.set_platform(args.device)
numpyro.set_host_device_count(args.num_chains)
# setup hyperparameters
hypers = {'expected_sparsity': max(1.0, args.num_dimensions / 2),
}
# add in hyperopt params
hypers.update(params)
# do inference
rng_key = random.PRNGKey(0)
samples = run_inference(model, args, rng_key, X, Y, hypers)
# compute the mean and square root variance of each coefficient theta_i
means, stds = vmap(lambda dim: analyze_dimension(samples, X, Y, dim, hypers))(np.arange(args.num_dimensions))
num_dims = len(means)
# print("Coefficients theta_1 to theta_%d used to generate the data:" % args.active_dimensions, expected_thetas)
# print("The single quadratic coefficient theta_{1,2} used to generate the data:", expected_pairwise)
active_dimensions = []
for dim, (mean, std) in enumerate(zip(means, stds)):
# we mark the dimension as inactive if the interval [mean - 3 * std, mean + 3 * std] contains zero
lower, upper = mean - 1.5 * std, mean + 1.5 * std
inactive = "inactive" if lower < 0.0 and upper > 0.0 else "active"
if inactive == "active":
active_dimensions.append(dim)
print("[dimension %02d/%02d] %s:\t%.2e +- %.2e" % (dim + 1, args.num_dimensions, inactive, mean, std))
print("Identified a total of %d active dimensions; expected %d." % (len(active_dimensions),
args.active_dimensions))
# Compute the mean and square root variance of coefficients theta_ij for i,j active dimensions.
# Note that the resulting numbers are only meaningful for i != j.
if len(active_dimensions) > 0:
dim_pairs = np.array(list(itertools.product(active_dimensions, active_dimensions)))
means, stds = vmap(lambda dim_pair: analyze_pair_of_dimensions(samples, X, Y,
dim_pair[0], dim_pair[1], hypers))(dim_pairs)
# print(dim_pairs)
dim_pair_arr = []
dim_pair_index = num_dims -1
dim_pair_name = []
for dim_pair, mean, std in zip(dim_pairs, means, stds):
dim1, dim2 = dim_pair
if dim1 >= dim2:
continue
dim_pair_index += 1
lower, upper = mean - 1.5 * std, mean + 1.5 * std
if not (lower < 0.0 and upper > 0.0):
dim_pair_arr.append(dim_pair_index)
dim_pair_name.append('%d and %d'%(dim1 + 1, dim2 + 1))
format_str = "Identified pairwise interaction between dimensions %d and %d: %.2e +- %.2e"
print(format_str % (dim1 + 1, dim2 + 1, mean, std))
active_dimensions = active_dimensions + dim_pair_arr
return np.exp(-len(active_dimensions))
from time import time
import jax.numpy as jnp
# https://pytorch.org/tutorials/recipes/recipes/benchmark.html
from jax import random
from numpyro import set_platform
from numpyro.distributions import Sine, SineSkewed
if __name__ == "__main__":
set_platform("gpu")
key = random.PRNGKey(0)
loc = jnp.array([0.0])
conc = jnp.array([1.0])
corr = jnp.array([0.6])
sine = Sine(loc, loc, conc, conc, corr)
skewness = jnp.array([0.3, -0.2])
ss = SineSkewed(sine, skewness)
first_timings = []
after_timings = []
for samples in [1, 5, 25, 125, 625, 3125, 15625, 78125, 390625, 1953125]:
times = []
print(samples)
for _ in range(11):
key, sample_key = random.split(key)
start = time()
data = ss.sample(sample_key, (samples, ))
times.append(time() - start)
jnp.savez(path, **params)
def load_params(path: Union[str, pathlib.Path]) -> Dict[str, jnp.ndarray]:
with jnp.load(path) as data:
res = {k: jnp.array(v) for k, v in data.items()}
return res
if __name__ == "__main__":
num_chains = 4
numpyro.set_platform("cpu")
numpyro.set_host_device_count(num_chains)
y = np.array([28.0, 8.0, -3.0, 7.0, -1.0, 1.0, 18.0, 12.0])
sigma = np.array([15.0, 10.0, 16.0, 11.0, 9.0, 11.0, 10.0, 18.0])
rng_key = random.PRNGKey(0)
rng_key, rng_key_posterior, rng_key_prior = random.split(rng_key, 3)
mcmc = inference(model, sigma, y, rng_key, num_chains=num_chains)
posterior_samples = mcmc.get_samples()
posterior_predictive = predict(model,
sigma,
rng_key_posterior,
posterior_samples=posterior_samples)
prior = predict(model, sigma, rng_key_prior, num_samples=500)
# Copyright 2019- d3p Developers and their Assignees
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import jax
import numpyro
import logging
try:
jax.lib.xla_bridge.get_backend(
'gpu') # this will fail if gpu not available
numpyro.set_platform('gpu')
except RuntimeError:
logging.info("GPU not available. Falling back to CPU.")
