Plot_reg_all

plot_reg_all.py

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import pandas as pd
from scipy import stats

TotalMSELoss = [0.2830, 0.4853, 0.6963, 1.0990, 1.2796, 1.7281, 1.8727, 2.0605, 2.1643, 2.5878, 2.7190, 3.0497, 3.1828, 3.3756, 3.4804,  3.9982, 4.6670]
NewLoss =      [1.3913, 1.4027, 1.4132, 1.4496, 1.4591, 1.4848, 1.4941, 1.5148, 1.5144, 1.5435, 1.5481, 1.5704, 1.5802, 1.5903, 1.5958,  1.6268, 1.6650]
Acc =          [75.52 , 75.00 , 74.66 , 74.76 , 74.46 , 74.36 , 74.43 , 74.57 , 74.52 , 74.55 , 74.34 , 73.80 , 74.33 , 74.14 , 73.86 ,  73.85 , 73.58]
MnasNet =      [49.74,     49.22,   48.13,  48.02,  47.76,  46.91,  46.68,  47.74,  47.77,  48.23,  47.61,  46.83,  47.90,  46.80,  47.11,  46.61,  46.11] # lr=0.001
MnasNet_2 =      [1.78,     16.01,  9.78,  12.83,  17.48,  14.24,  15.73,  16.12,  13.36,  14.28,  14.97,  13.89,  13.57,  15.39,  15.03,  15.83,  13.47] # lr=0.2
SuperNetInfer= [70.80 , 70.86 , 70.768, 70.704, 70.896, 70.796, 70.558, 70.62 , 70.892, 70.86 , 70.804, 70.848, 70.734, 70.918, 71.024,  70.746, 70.846 ]


TotalMSELoss = np.array(TotalMSELoss)
NewLoss = np.array(NewLoss)
MnasNet = np.array(MnasNet)
MnasNet_2 = np.array(MnasNet_2)
Acc = np.array(Acc)
SuperNetInfer = np.array(SuperNetInfer)


DataDict = {"Stand-Alone Model Accuracy": Acc, "Total Evaluation Loss (minus)": -NewLoss, 'Undertrained(0.001)': MnasNet, 'Undertrained(0.2)': MnasNet_2 , 'SPOS': SuperNetInfer}
DataFrame = pd.DataFrame(DataDict)

def set_axes(ax1, y1, y2, bias=(0,1,0.1)):
    ax2 = ax1.twinx()
    y1_data, y2_data = DataDict[y1], DataDict[y2]
    print(y1_data, y2_data)
    ax1.set_ylim([np.min(y1_data)-bias[0], np.max(y1_data)+bias[0]])
    ax2.set_ylim([np.min(y2_data)-bias[1], np.max(y2_data)+bias[1]])
    return ax1, ax2

def plot_data(data, plot_cfg):

    x_axis = "Stand-Alone Model Accuracy"
    y_axis = [("Total Evaluation Loss (minus)", "DNA"),
              ("Undertrained(0.001)", "MnasNet_0.001"), ("Undertrained(0.2)", "MnasNet_0.2"), ("SPOS", "SPOS")]

    print("Cal kendall tau")
    dnatau = stats.kendalltau(DataDict[x_axis], DataDict[y_axis[0][0]])
    print("DNA :{}".format(dnatau))
    mnasNet1tau = stats.kendalltau(DataDict[x_axis], DataDict[y_axis[1][0]])
    print("MnasNet 0.001 :{}".format(mnasNet1tau))
    mnasNet2tau = stats.kendalltau(DataDict[x_axis], DataDict[y_axis[2][0]])
    print("MnasNet 0.2 :{}".format(mnasNet2tau))
    spostau = stats.kendalltau(DataDict[x_axis], DataDict[y_axis[3][0]])
    print("MnasNet 0.2 :{}".format(spostau))

    sns.set(style='white', palette='muted', color_codes=True)
    sns.set(style='white')
    colors = sns.color_palette('Set1', n_colors=5)
    # using axes
    f, axes = plt.subplots(1,4,figsize=(32,8))
    # plt.subplots_adjust(bottom=0.25, top=0.615, wspace=0.2, right=0.95, left=0.05)

    ci = 98
    fontsize = 25
    axis_fsize = 18

    ax1,ax2,ax3,ax4 = axes[0], axes[1], axes[2], axes[3]

    ax11, ax12 = set_axes(ax1, y_axis[0][0], y_axis[1][0], bias=(0.02, 0.1))

    sns.regplot(x = x_axis, y = y_axis[0][0],  data = DataFrame, ax = ax11, label = y_axis[0][1]+'({0:.2f})'.format(dnatau.correlation), scatter_kws={'color':colors[0]}, line_kws={"color": colors[0]}, ci=ci)

    ax21, ax22 = set_axes(ax2, y_axis[0][0], y_axis[3][0], bias=(0.02, 0.1))
    sns.regplot(x = x_axis, y = y_axis[0][0],  data = DataFrame, ax = ax21, scatter_kws={'color':colors[0]}, line_kws={"color": colors[0]}, ci=ci)
    sns.regplot(x = x_axis, y = y_axis[3][0],  data = DataFrame, ax = ax22, label = y_axis[3][1]+'({0:.2f})'.format(spostau.correlation),  scatter_kws={'color': colors[4]}, line_kws={"color": colors[4]}, ci=ci)
    ax21.set_title('v.s. One-shot Based', fontsize=fontsize)
    ax21.set_xlabel("")
    ax21.set_ylabel("")
    ax22.set_xlabel("")
    ax22.set_ylabel("")
    ax21.tick_params(labelsize=axis_fsize)
    ax22.tick_params(labelsize=axis_fsize)

    ax31, ax32 = set_axes(ax3, y_axis[0][0], y_axis[2][0], bias=(0.02, 0.1))
    sns.regplot(x = x_axis, y = y_axis[0][0],  data = DataFrame, ax = ax31, scatter_kws={'color':colors[0]}, line_kws={"color": colors[0]}, ci=ci)
    sns.regplot(x = x_axis, y = y_axis[2][0],  data = DataFrame, ax = ax32, label = y_axis[2][1]+'({0:.2f})'.format(mnasNet2tau.correlation),  scatter_kws={'color': colors[3]}, line_kws={"color": colors[3]}, ci=ci)
    ax31.set_title('v.s. Undertrained(0.2)', fontsize=fontsize)
    ax31.set_xlabel("")
    ax31.set_ylabel("")
    ax32.set_xlabel("")
    ax32.set_ylabel("")
    ax31.tick_params(labelsize=axis_fsize)
    ax32.tick_params(labelsize=axis_fsize)

    ax41, ax42 = set_axes(ax4, y_axis[0][0], y_axis[1][0], bias=(0.02, 0.1))
    sns.regplot(x = x_axis, y = y_axis[0][0],  data = DataFrame, ax = ax41, scatter_kws={'color':colors[0]}, line_kws={"color": colors[0]}, ci=ci)
    sns.regplot(x = x_axis, y = y_axis[1][0],  data = DataFrame, ax = ax42, label = y_axis[1][1]+'({0:.2f})'.format(mnasNet1tau.correlation),  scatter_kws={'color': colors[2]}, line_kws={"color": colors[2]}, ci=ci)
    ax41.set_title('v.s. Undertrained(0.001)',fontsize=fontsize)
    ax41.set_xlabel("")
    ax41.set_ylabel("")
    ax42.set_xlabel("")
    ax42.set_ylabel("")
    ax41.tick_params(labelsize=axis_fsize)
    ax42.tick_params(labelsize=axis_fsize)



    f.legend(loc=(0.87,0.08), fontsize=16)
    plt.xticks(fontsize=axis_fsize)
    plt.yticks(fontsize=axis_fsize)
    # plt.setp(ax1.get_xticklabels(), visible=True)
    f.tight_layout()
    plt.savefig('./Allloss.pdf')


    plt.show()


if __name__ == '__main__':
    plot_data(data=None, plot_cfg=None)