def plot_convergence_iterations_for_data(parsed_data, exp_num):
convergence_data, _ = Parser.get_convergence_and_distinct_patterns_from_log_v1(parsed_data)
# for all data: 4 buckets, containing 4x100 data points.
# convergence_stats = Parser.get_convergence_avgs(convergence_data)
ds = []
for i in range(4):
ds.append([convergence_data[0][i*100:i*100 + 100], convergence_data[1][i*100:i*100 + 100],
convergence_data[2][i*100:i*100 + 100], convergence_data[3][i*100:i*100 + 100]])
average_iters_convergence, sigmas = Parser.get_average_recall_ratios(ds[exp_num])
# Plotting:
x = np.asarray([2, 3, 4, 5])
plt.rcParams.update({'font.size': 25})
p1 = plt.plot(x, average_iters_convergence, color='b', marker='o', linestyle='--', linewidth=3.0)
p2 = plt.plot(x, sigmas, color='r', marker='^', linestyle='--', linewidth=3.0)
# p1 = plt.plot(x, [2.78, 19.02, 15.85, 15.58], color='r', marker='o', linestyle='--')
plt.ylabel('Average #iterations before convergence')
plt.xlabel('Set size')
plt.title('Average #iterations before convergence by set size')
plt.xticks(x, ('2', '3', '4', '5'))
plt.legend((p1[0], p2[0]), ('Sync., turnover for every learnt set', 'Standard deviation'))
# bbox_to_anchor=(0.445, 1))
plt.grid(True)
plt.show()
def plot_convergence_ratios_for_data(parsed_data, data_points_per_experiment_config):
convergence_data, _ = Parser.get_convergence_and_distinct_patterns_from_log_v1(parsed_data)
# for all data: 4 buckets, containing 4x100 data points.
# convergence_stats = Parser.get_convergence_avgs(convergence_data)
ds = []
for i in range(4):
ds.append([convergence_data[0]
[i * data_points_per_experiment_config:i * data_points_per_experiment_config +
data_points_per_experiment_config],
convergence_data[1][i * data_points_per_experiment_config:i * data_points_per_experiment_config +
data_points_per_experiment_config],
convergence_data[2][i * data_points_per_experiment_config:i * data_points_per_experiment_config +
data_points_per_experiment_config],
convergence_data[3][i * data_points_per_experiment_config:i * data_points_per_experiment_config +
data_points_per_experiment_config]])
convergence_stats_async_true_mode_0 = Parser.get_convergence_avgs(ds[0])[1]
convergence_stats_async_true_mode_1 = Parser.get_convergence_avgs(ds[1])[1]
convergence_stats_async_false_mode_0 = Parser.get_convergence_avgs(ds[2])[1]
convergence_stats_async_false_mode_1 = Parser.get_convergence_avgs(ds[3])[1]
# Plotting:
x = np.asarray([2, 3, 4, 5])
plt.rcParams.update({'font.size': 25})
print "y:", convergence_stats_async_true_mode_0
# print "e:", e1.shape
# print "e:", e1
p1 = plt.plot(x, convergence_stats_async_true_mode_0, color='y', marker='o', linestyle='--', linewidth=3.0)
p2 = plt.plot(x, convergence_stats_async_true_mode_1, color='g', marker='s', linestyle='--', linewidth=3.0)
p3 = plt.plot(x, convergence_stats_async_false_mode_0, color='b', marker='^', linestyle='--', linewidth=3.0)
p4 = plt.plot(x, convergence_stats_async_false_mode_1, color='c', marker='o', linestyle='--', linewidth=3.0)
plt.ylabel('Average convergence rate (%)')
plt.xlabel('Set size')
plt.title('Average convergence rate by set size')
plt.xticks(x, ('2x5', '3x5', '4x5', '5x5'))
plt.legend((p1[0], p2[0], p3[0], p4[0]), ('Async. tm=0', 'Async. tm=1',
'Sync. tm=0', 'Sync. tm=1'),
bbox_to_anchor=(0.27, 0.28))
plt.margins(0.025)
plt.grid(True)
plt.show()
def plot_pattern_stats_from_parsed_data_v1(parsed_data, exp_num):
_, distinct_patterns_data = Parser.get_convergence_and_distinct_patterns_from_log_v1(parsed_data)
perfect_recall_rates, spurious_patterns_extracted = Parser. \
get_perfect_recall_rates_and_spurious_patterns_from_data(parsed_data)
dist_patterns_data_experiments = []
pf_buckets = []
spurious_buckets = []
for i in range(4):
dist_patterns_data_experiments.append([])
pf_buckets.append([])
spurious_buckets.append([])
for i in range(4):
dist_patterns_data_experiments[i].append(distinct_patterns_data[0][i*100: i*100 + 100])
dist_patterns_data_experiments[i].append(distinct_patterns_data[1][i*100: i*100 + 100])
dist_patterns_data_experiments[i].append(distinct_patterns_data[2][i*100: i*100 + 100])
dist_patterns_data_experiments[i].append(distinct_patterns_data[3][i*100: i*100 + 100])
pf_buckets[i].append(perfect_recall_rates[0][i*20: i*20 + 20])
pf_buckets[i].append(perfect_recall_rates[1][i*20: i*20 + 20])
pf_buckets[i].append(perfect_recall_rates[2][i*20: i*20 + 20])
pf_buckets[i].append(perfect_recall_rates[3][i*20: i*20 + 20])
spurious_buckets[i].append(spurious_patterns_extracted[0][i*20: i*20 + 20])
spurious_buckets[i].append(spurious_patterns_extracted[1][i*20: i*20 + 20])
spurious_buckets[i].append(spurious_patterns_extracted[2][i*20: i*20 + 20])
spurious_buckets[i].append(spurious_patterns_extracted[3][i*20: i*20 + 20])
# convergence_data[0]: convergence data for the first set size.
avg_recall, stds_avg_recall = Parser.get_average_recall_ratios(dist_patterns_data_experiments[exp_num])
for i in range(len(spurious_buckets[exp_num])):
for j in range(len(spurious_buckets[exp_num][i])):
spurious_buckets[exp_num][i][j] = spurious_buckets[exp_num][i][j] / 5.
# print spurious_buckets[0][1]
avg_spurious_ratios, stds_spurious = Parser.get_average_recall_ratios(spurious_buckets[exp_num])
# print avg_recall
# print "stds_avg_recall:", stds_avg_recall
# print
# print avg_spurious_ratios
# print "stds_spurious:", stds_spurious
# graph 1: box plot
width = .35
V = np.arange(4)
p1 = plt.bar(V, avg_recall, width=width, color='y', yerr=stds_avg_recall)
p2 = plt.bar(V + width, avg_spurious_ratios, width=width, color='r', yerr=stds_spurious)
plt.rcParams.update({'font.size': 25})
plt.ylabel('Number of patterns')
plt.xlabel('Set size')
plt.title('Average number of patterns recalled for each sub-set')
plt.xticks(V + width, ('2', '3', '4', '5'))
plt.yticks(np.arange(0, 6, 1))
plt.legend((p1[0], p2[0]), ('avg. # patterns recalled per sub-set', 'avg. spurious recall per sub-set'),
bbox_to_anchor=(0.42, 1))
plt.grid(True)
plt.show()
