riskFnGauss.ipynb

Kernel: Unknown Kernel

# Convert riskFnGauss(Fig 5.13) to python/JAX
# Author: Garvit9000c


import numpy as np
import matplotlib.pyplot as plt

for n in [5, 20]:
    # Domain
    mus = np.arange(-1.8, 1.8, 0.1)

    # constants
    M = len(mus)
    n0 = 1
    n0B = 5
    mu0 = 0
    pi = np.pi

    # functions
    r1 = (1 / n) * np.ones(M)
    r2 = (pi / (2 * n)) * np.ones(M)
    r3 = (mus - mu0) ** 2
    r4 = (n + n0) ** (-2) * (n + n0**2 * (mu0 - mus) ** 2)
    r5 = (n + n0B) ** (-2) * (n + n0B**2 * (mu0 - mus) ** 2)

    # Plotting
    plt.title("risk function for n=" + str(n))
    legendStr = ["mle", "median", "fixed", "postmean1", "postmean5"]
    style = [["b", "solid"], ["r", "dotted"], ["k", "dashdot"], ["lime", "dashed"], ["cyan", "solid"]]

    r = [r1, r2, r3, r4, r5]
    for i in range(len(r)):
        plt.plot(mus, r[i], style[i][0], linestyle=style[i][1], label=legendStr[i])
    plt.legend(loc="upper left", prop={"size": 7})

    if n == 5:
        plt.ylim(0, 0.5)
    else:
        plt.ylim(0, 0.18)
    plt.xticks([-2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2])

    # axis Labels
    plt.xlabel(r"$\theta_*$")
    plt.ylabel(r"$R(\theta_*,\delta)$")

    D = {5: "a", 20: "b"}
    plt.savefig("figures/riskFnGauss(" + D[n] + ").pdf", dpi=300)
    plt.show()