具有伯努利分布的 TensorFlow Probability MCMC

Question

我需要使用 TensorFlow Probability 从伯努利分布中采样来实现马尔可夫链蒙特卡罗。但是，我的尝试显示的结果与我对伯努利分布的期望不一致。我在这里修改了 tfp.mcmc.sample_chain（从对角线方差高斯采样）示例的文档中给出的示例，以从伯努利分布中提取。由于伯努利分布是离散的，我使用了 RandomWalkMetropolis 转换内核而不是 Hamiltonian Monte Carlo 内核，我预计它不会工作，因为它计算梯度。这是代码：import numpy as npimport matplotlib.pyplot as pltimport seaborn as snsimport tensorflow as tfimport tensorflow_probability as tfptfd = tfp.distributionsdef make_likelihood(event_prob):    return tfd.Bernoulli(probs=event_prob,dtype=tf.float32)dims=1event_prob = 0.3num_results = 30000likelihood = make_likelihood(event_prob)states, kernel_results = tfp.mcmc.sample_chain(    num_results=num_results,    current_state=tf.zeros(dims),    kernel = tfp.mcmc.RandomWalkMetropolis(              target_log_prob_fn=likelihood.log_prob,              new_state_fn=tfp.mcmc.random_walk_normal_fn(scale=1.0),              seed=124             ),    num_burnin_steps=5000)chain_vals = states# Compute sample stats.sample_mean = tf.reduce_mean(states, axis=0)sample_var = tf.reduce_mean(    tf.squared_difference(states, sample_mean),    axis=0)#initialize the variableinit_op = tf.global_variables_initializer()#run the graphwith tf.Session() as sess:    sess.run(init_op)     [sample_mean_, sample_var_, chain_vals_] = sess.run([sample_mean,sample_var,chain_vals])chain_samples = (chain_vals_[:] )   print ('Sample mean = {}'.format(sample_mean_))print ('Sample var = {}'.format(sample_var_))fig, axes = plt.subplots(2, 1)fig.set_size_inches(12, 10)axes[0].plot(chain_samples[:])axes[0].title.set_text("values sample chain tfd.Bernoulli")sns.kdeplot(chain_samples[:,0], ax=axes[1], shade=True)axes[1].title.set_text("chain tfd.Bernoulli distribution")fig.tight_layout()plt.show()我希望看到区间 [0,1] 中马尔可夫链状态的值。马尔可夫链值的结果看起来不像伯努利分布的预期结果，KDE 图也不是，如下图所示：我的示例是否存在概念上的缺陷，或者在使用 TensorFlow Probability API 时是否存在错误？或者使用离散分布（例如伯努利分布）的马尔可夫链蒙特卡罗的 TF.Probability 实现可能存在问题？