def __init__(self,
cases=[1],
K=[5],
n_total=[1000],
n_samples=[[5, 5]],
b_true=[None],
w_true=[None],
num_results=[10e3],
baseline_index=None,
formula="x_0"):
"""
constructor. Simulated Parameters are passed to the constructor as lists, except the type of model, which is fixed for all simulations.
The simulation is carried out over all possible combinations of specified parameters.
See the default values for examples
Parameters
----------
cases -- List
Number of (binary) covariates
K -- List
Number of cell types
n_total -- List
number of cells per sample
n_samples -- List
number of samples. Each sublist specifies the number of samples for each covariate combination, length 2**cases
b_true -- List
Base composition. Each sublist has dimension K.
w_true -- List
Effect composition. Each sublist is a nested list that represents a DxK effect matrix
num_results -- List
MCMC chain length
baseline_index -- int
Index of reference cellltype (None for no baseline)
formula -- str
R-style formula used in model specification
"""
# HMC Settings
self.n_burnin = int(5e3) # number of burn-in steps
self.step_size = 0.01
self.num_leapfrog_steps = 10
# All parameter combinations
self.simulation_params = list(
itertools.product(cases, K, n_total, n_samples, b_true, w_true,
num_results))
# Setup result objects
self.mcmc_results = {}
self.parameters = pd.DataFrame({
'cases': [],
'K': [],
'n_total': [],
'n_samples': [],
'b_true': [],
'w_true': [],
'num_results': []
})
self.baseline_index = baseline_index
self.formula = formula
np.set_seed(1234)
# limitations under the License.
from __future__ import print_function
import unittest
import numpy
import paddle.fluid as fluid
import paddle.fluid.layers as layers
import paddle.fluid.core as core
from paddle.fluid.contrib.layers import BasicLSTMUnit
from paddle.fluid.executor import Executor
from paddle.fluid import framework
import numpy as np
np.set_seed(123)
SIGMOID_THRESHOLD_MIN = -40.0
SIGMOID_THRESHOLD_MAX = 13.0
EXP_MAX_INPUT = 40.0
def sigmoid(x):
y = np.copy(x)
y[x < SIGMOID_THRESHOLD_MIN] = SIGMOID_THRESHOLD_MIN
y[x > SIGMOID_THRESHOLD_MAX] = SIGMOID_THRESHOLD_MAX
return 1. / (1. + np.exp(-y))
def tanh(x):
y = -2. * x
# %%
plt.plot()
film_deaths.IMDB_Rating.hist(bins=10)
plt.xlabel('IMBD_Rating')
plt.ylabel('Films Count')
plt.show()
# %%
imdb_mean = film_deaths.IMDB_Rating.mean()
imdb_std = film_deaths.IMDB_Rating.std()
print(f'IMDB mean = {imdb_mean:.3}, std = {imdb_std:.3}')
# %%
np.set_seed(42)
film_deaths['imdb_simulation'] = np.random.normal(loc=imdb_mean,
scale=imdb_std,
size=len(film_deaths))
plt.plot()
film_deaths.imdb_simulation.hist(bins=10)
plt.xlabel('IMDB_Rating (Simulation)')
plt.ylabel('Films Count (Simulation)')
plt.show()
# %%
from scipy.stats import probplot