我正在尝试在张量流概率中拟合一个简单的 Dirichlet-Multinomial 模型。浓度参数是gamma
并且我已经在它们上放置了 Gamma(1,1) 先验分布。这是模型,其中 S 是类别数,N 是样本数:
def dirichlet_model(S, N):
gamma = ed.Gamma(tf.ones(S)*1.0, tf.ones(S)*1.0, name='gamma')
y = ed.DirichletMultinomial(total_count=500., concentration=gamma, sample_shape=(N), name='y')
return y
log_joint = ed.make_log_joint_fn(dirichlet_model)
但是,当我尝试使用 HMC 从中采样时,接受率为零,并且初始抽签gamma
包含负值。难道我做错了什么?不应该自动拒绝对浓度参数的否定建议吗?在我的采样代码下面:
def target_log_prob_fn(gamma):
"""Unnormalized target density as a function of states."""
return log_joint(
S=S, N=N,
gamma=gamma,
y=y_new)
num_results = 5000
num_burnin_steps = 3000
states, kernel_results = tfp.mcmc.sample_chain(
num_results=num_results,
num_burnin_steps=num_burnin_steps,
current_state=[
tf.ones([5], name='init_gamma')*5,
],
kernel=tfp.mcmc.HamiltonianMonteCarlo(
target_log_prob_fn=target_log_prob_fn,
step_size=0.4,
num_leapfrog_steps=3))
gamma = states
with tf.Session() as sess:
[
gamma_,
is_accepted_,
] = sess.run([
gamma,
kernel_results.is_accepted,
])
num_accepted = np.sum(is_accepted_)
print('Acceptance rate: {}'.format(num_accepted / num_results))