''' ========================= Plotting Results ========================= Code example to plot the results after the experiment. ''' try: import cPickle as pickle except: import pickle import matplotlib import matplotlib.pyplot as plt import itertools from matplotlib import rc import numpy as np import pandas as pd import seaborn font = {'family': 'serif', 'serif': ['computer modern roman']} rc('text', usetex=False) rc('font', weight='bold') rc('font', size=8) rc('lines', markersize=2.5) rc('lines', linewidth=0.5) rc('xtick', labelsize=6) rc('ytick', labelsize=6) rc('axes', labelsize='small') rc('axes', labelweight='bold') rc('axes', titlesize='small') rc('axes', linewidth=1) plt.rc('font', **font) seaborn.set_style("darkgrid") import pdb def plot_benchmark(domains, graph_attrs, attr_defaults, methods, graph_names, s_sch='rw'): df_all = pd.DataFrame() for d, m , g in itertools.product(*[domains, methods, graph_names]): try: df = pd.read_hdf(path+ "%s_%s_%s_lp_%s_data_hyp.h5" % (d, g, m, s_sch), "df" ) except: print('%s_%s_%s_lp_%s_data_hyp.h5 not found. Ignoring data set' % (d, g, m, s_sch)) continue df["Domain"], df["Method"], df["Graph"] = d, m, g df_all = df_all.append(df).reset_index() df_all = df_all.drop(['index'], axis=1) if df_all.empty: return df_all = df_all.drop(['dia'], axis=1) plot_shape = (len(domains), len(graph_attrs)) fin1, axarray1 = plt.subplots(len(domains), len(graph_attrs), figsize=(7, 4), sharex='col', sharey='row') data_idx = 0 gfs_score = {} for dom in domains: gfs_score[dom] = {m: 0 for m in methods} n_attr = 0 for attr in graph_attrs: plot_idx = np.unravel_index(data_idx, plot_shape) data_idx += 1 try: rem_attrs = list(set(graph_attrs) - {attr}) df_grouped = df_all[df_all["Domain"]==dom] for rem_attr in rem_attrs: df_grouped = df_grouped[df_grouped[rem_attr]==attr_defaults[rem_attr]] df_grouped = df_grouped[[attr, "Round Id", "LP MAP", "Method", "Graph"]] print(df_grouped.head()) df_grouped['LP MAP'] = df_grouped['LP MAP'].astype('float') df_grouped = df_grouped.groupby([attr, "Round Id", "Method", "Graph"]).mean().reset_index() except TypeError: df_trun[hyp_key_ren + "2"] = \ df_trun[hyp_key_ren].apply(lambda x: str(x)) df_trun[hyp_key_ren] = df_trun[hyp_key_ren + "2"].copy() df_trun = df_trun.drop([hyp_key_ren + "2"], axis=1) df_grouped = df_trun.groupby([hyp_key_ren, "Round Id", "Data"]).max().reset_index() df_grouped['unit']=df_grouped.apply(lambda x:'%s_%s' % (x['Round Id'],x['Graph']),axis=1) df_grouped = df_grouped.drop(['Round Id', "Graph"], axis=1) m_scores = dict(df_grouped.groupby(["Method"])["LP MAP"].mean()) n_attr += 1 gfs_score[dom] = {m: m_scores[m] + gfs_score[dom][m] for m in methods} ax = seaborn.tsplot(time=attr, value="LP MAP", unit="unit", condition="Method", legend=False, data=df_grouped, ax=axarray1[plot_idx[0], plot_idx[1]]) if not plot_idx[1]: ax.set_ylabel(dom) if plot_idx[1]: ax.set_ylabel('') if plot_idx[0] < len(domains) - 1: ax.set_xlabel('') attr_values = df_grouped[attr].unique() l_diff = attr_values[-1] - attr_values[-2] f_diff = attr_values[1] - attr_values[0] l_f_diff_r = l_diff / f_diff if l_f_diff_r > 1: log_base = pow(l_f_diff_r, 1.0 / (len(attr_values) - 2)) ax.set_xscale('log', basex=round(log_base)) marker = ["o", "s", "D", "^", "v", "8", "*", "p", "1", "h"] for line_i in range(len(ax.lines)): ax.lines[line_i].set_marker(marker[line_i]) gfs_score[dom] = {m: gfs_score[dom][m]/n_attr for m in methods} print(gfs_score) plt.savefig( 'benchmark.pdf', dpi=300, format='pdf', bbox_inches='tight' ) plt.show() def plot_benchmark_individual(domains, graph_attrs, attr_defaults, methods, graph_names, graph_short, s_sch='rw'): df_all = pd.DataFrame() for d, m , g in itertools.product(*[domains, methods, graph_names]): try: df = pd.read_hdf(path+ "%s_%s_%s_lp_%s_data_hyp.h5" % (d, g, m, s_sch), "df" ) except: print('%s_%s_%s_lp_%s_data_hyp.h5 not found. Ignoring data set' % (d, g, m, s_sch)) continue df["Method"], df["Graph"] = m, g df_all = df_all.append(df).reset_index() df_all = df_all.drop(['index'], axis=1) if df_all.empty: return df_all = df_all.drop(['dia'], axis=1) plot_shape = (len(graph_names), len(graph_attrs)) fin1, axarray1 = plt.subplots(len(graph_names), len(graph_attrs), figsize=(7, 4), sharex='col', sharey='row') data_idx = 0 for graph in graph_names: for attr in graph_attrs: plot_idx = np.unravel_index(data_idx, plot_shape) data_idx += 1 try: rem_attrs = list(set(graph_attrs) - {attr}) df_grouped = df_all[df_all["Graph"]==graph] for rem_attr in rem_attrs: df_grouped = df_grouped[df_grouped[rem_attr]==attr_defaults[rem_attr]] df_grouped = df_grouped[[attr, "Round Id", "LP MAP", "Method"]] print(df_grouped.head()) df_grouped['LP MAP'] = df_grouped['LP MAP'].astype('float') df_grouped = df_grouped.groupby([attr, "Round Id", "Method"]).mean().reset_index() except TypeError: df_trun[hyp_key_ren + "2"] = \ df_trun[hyp_key_ren].apply(lambda x: str(x)) df_trun[hyp_key_ren] = df_trun[hyp_key_ren + "2"].copy() df_trun = df_trun.drop([hyp_key_ren + "2"], axis=1) df_grouped = df_trun.groupby([hyp_key_ren, "Round Id", "Data"]).max().reset_index() if data_idx == len(graph_names) * len(graph_attrs): legend = True else: legend = False ax = seaborn.tsplot(time=attr, value="LP MAP", unit="Round Id", condition="Method", legend=legend, data=df_grouped, ax=axarray1[plot_idx[0], plot_idx[1]]) if legend: ax.legend_.remove() if not plot_idx[1]: ax.set_ylabel(graph_short[graph]) if plot_idx[1]: ax.set_ylabel('') if plot_idx[0] < len(graph_names) - 1: ax.set_xlabel('') attr_values = df_grouped[attr].unique() l_diff = attr_values[-1] - attr_values[-2] f_diff = attr_values[1] - attr_values[0] l_f_diff_r = l_diff / f_diff if l_f_diff_r > 1: log_base = pow(l_f_diff_r, 1.0 / (len(attr_values) - 2)) ax.set_xscale('log', basex=round(log_base)) marker = ["o", "s", "D", "^", "v", "8", "*", "p", "1", "h"] for line_i in range(len(ax.lines)): ax.lines[line_i].set_marker(marker[line_i]) for col_idx in range(axarray1[3].shape[0]): box = axarray1[3][col_idx].get_position() axarray1[3][col_idx].set_position( [box.x0, box.y0 + box.height * 0.1, box.width, box.height * 0.9] ) fin1.legend(loc='lower center', bbox_to_anchor=(0.45, -0.01), ncol=len(methods), fancybox=True, shadow=True) fin1.savefig( 'benchmark_individual3.pdf', dpi=300, format='pdf', bbox_inches='tight' ) plt.show() #Path for the stored .h5 result path = '/Users/Bench_files/new_files/' plot_benchmark( ["social", "biology", "internet"], ["N", "dim", "deg"], {"N": 4096, "deg": 8, "dia": None, "dim": 128}, ["gf", "rand", "pa", "lap", "hope", "cn", "aa","sdne"], ["watts_strogatz_graph", "barabasi_albert_graph", "powerlaw_cluster_graph", "random_geometric_graph",\ "duplication_divergence_graph","hyperbolic_graph","r_mat_graph","waxman_graph","stochastic_block_model",\ "stochastic_kronecker_graph"], s_sch='rw' ) plot_benchmark_individual( ["social", "biology", "internet"], ["N", "dim", "deg"], {"N": 4096, "deg": 8, "dia": None, "dim": 128}, ["gf", "rand", "pa", "lap", "hope", "cn", "aa", "sdne"], ["watts_strogatz_graph", "barabasi_albert_graph", "powerlaw_cluster_graph", "random_geometric_graph",\ "duplication_divergence_graph","hyperbolic_graph","r_mat_graph","waxman_graph","stochastic_kronecker_graph"], {"watts_strogatz_graph": "WS", "barabasi_albert_graph": "BA", \ "powerlaw_cluster_graph": "PC", "random_geometric_graph": "RG",\ "duplication_divergence_graph": "DD", "hyperbolic_graph": "HB",\ "r_mat_graph": "RM", "waxman_graph":"WM","stochastic_block_model":"SBM", "stochastic_kronecker_graph": "KG" }, s_sch='rw' )