Factor Metropolis X2X4 patch.py

From Werner KRAUTH

Revision as of 07:00, 11 June 2024 Werner (Talk | contribs) ← Previous diff		Current revision Werner (Talk | contribs)
Line 45:		Line 45:
	==Further information==		==Further information==
	==References==		==References==
		+	* Tartero, G., Krauth, W. Concepts in Monte Carlo sampling, Am. J. Phys. 92, 65–77 (2024) [https://arxiv.org/pdf/2309.03136 arXiv:2309.03136

---

Current revision

Contents 1 Context 2 Python program 3 Further information 4 References

Context

This page is part of my 2024 Beg Rohu Lectures on "The second Markov chain revolution" at the Summer School "Concepts and Methods of Statistical Physics" (3 - 15 June 2024).

Python program

import math
import random
import matplotlib.pyplot as plt

def u(x): return x ** 2 / 2.0 + x ** 4 / 4.0

def u2(pos): return 0.5 * pos ** 2

def u4(pos): return 0.25 * pos ** 4

x = 0.0
delta = 0.1

data = []
n_samples = 10 ** 6
for i in range(n_samples):
    new_x = x + random.uniform(-delta, delta)
    delta_u2 = u2(new_x) - u2(x)
    delta_u4 = u4(new_x) - u4(x)
    Upsilon2 = random.uniform(0.0, 1.0)
    Upsilon4 = random.uniform(0.0, 1.0)
    if Upsilon2 < math.exp(-delta_u2) and Upsilon4 < math.exp(-delta_u4):
        x = new_x
    data.append(x)
plt.title('Factorized Metropolis algorithm (patch), anharmonic oscillator' )
plt.xlabel('$x$')
plt.ylabel('$\pi(x)$')
plt.hist(data, bins=100, density=True,label='data')
XValues = []
YValues = []
for i in range(-1000,1000):
    x = i / 400.0
    XValues.append(x)
    YValues.append(math.exp(- u(x)) / 1.93525)
plt.plot(XValues, YValues, label='theory')
plt.legend(loc='upper right')
plt.show()

Further information

References

• Tartero, G., Krauth, W. Concepts in Monte Carlo sampling, Am. J. Phys. 92, 65–77 (2024) [https://arxiv.org/pdf/2309.03136 arXiv:2309.03136