# %% Import packages from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path # %% Define optimizer-specific output directories optimizer_output_path = output_path.joinpath('sgd') optimizer_output_pilot_path = optimizer_output_path.joinpath('pilot_run') optimizer_output_benchmark_path = optimizer_output_path.joinpath('benchmark_run')
# %% Import packages from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.mcmc.constants import num_chains # %% Define sampler-specific output directories sampler_output_path = output_path.joinpath('prior') sampler_output_run_paths = [ sampler_output_path.joinpath('run' + str(i + 1).zfill(len(str(num_chains)))) for i in range(num_chains) ]
# %% Load packages import pandas as pd from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.stats import mc_efficiency # %% Define sampler-specific output directories sampler_output_paths = output_path.joinpath('metropolis_hastings') sampler_output_paths = [ output_path.joinpath(name) for name in ['metropolis_hastings', 'mala', 'smmala'] ] # %% Compute Monte Carlo efficiency efficiency = mc_efficiency(sampler_output_paths, keys=['rhat', 'ess']) # %% Save Monte Carlo efficiency df = pd.DataFrame(data=efficiency) df.to_csv(output_path.joinpath('mc_efficiency.csv'), index=False) with open(output_path.joinpath('mc_efficiency.tex'), 'w') as file: file.write(df.round({'rhat': 4}).to_latex(index=False))
# %% Load packages import matplotlib.pyplot as plt import numpy as np import seaborn as sns from kanga.plots import redblue_cmap from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.mcmc.constants import pred_interval_x1, pred_interval_x2 # %% Load ground truth pred_interval_y = np.loadtxt(output_path.joinpath('ground_truth.csv'), delimiter=',', skiprows=0) # %% Plot heat map of ground truth num_ticks = 8 xticks = np.linspace(0, len(pred_interval_x1) - 1, num=num_ticks, dtype=np.int) xticklabels = [np.round(pred_interval_x1[idx], decimals=2) for idx in xticks] yticks = np.linspace(0, len(pred_interval_x2) - 1, num=num_ticks, dtype=np.int) yticklabels = [np.round(pred_interval_x2[idx], decimals=2) for idx in yticks] ax = sns.heatmap(pred_interval_y, cmap=redblue_cmap, linewidths=0.01, linecolor='white',
# %% Import packages from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.mcmc.constants import num_chains # %% Define sampler-specific output directories sampler_output_path = output_path.joinpath('metropolis_hastings') sampler_output_pilot_path = sampler_output_path.joinpath('pilot_run') sampler_output_run_paths = [ sampler_output_path.joinpath('run' + str(i + 1).zfill(len(str(num_chains)))) for i in range(num_chains) ]
# %% Import packages from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.mcmc.constants import num_chains # %% Define sampler-specific output directories sampler_output_path = output_path.joinpath('hmc') sampler_output_pilot_path = sampler_output_path.joinpath('pilot_run') sampler_output_run_paths = [ sampler_output_path.joinpath('run' + str(i + 1).zfill(len(str(num_chains)))) for i in range(num_chains) ]
# %% Import packages from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.constants import output_path from bnn_mcmc_examples.examples.mlp.noisy_xor.setting1.mcmc.constants import num_chains # %% Define sampler-specific output directories sampler_output_path = output_path.joinpath('power_posteriors') sampler_output_pilot_path = sampler_output_path.joinpath('pilot_run') sampler_output_run_paths = [ sampler_output_path.joinpath('run' + str(i + 1).zfill(len(str(num_chains)))) for i in range(num_chains) ]