Normalizing Flows

%matplotlib inline
from echoflow.demo import load_dataset
import matplotlib.pyplot as plt

df = load_dataset()
plt.figure(figsize=(6,6))
plt.scatter(df["x"], df["y"])

<matplotlib.collections.PathCollection at 0x7fd32c781400>

from echoflow import EchoFlow

model = EchoFlow(nb_blocks=10, block_type="RNVP", nb_epochs=100)
model.fit(df)
synthetic = model.sample(num_samples=1000)
plt.scatter(synthetic["x"], synthetic["y"])

Epoch 10 | Train Loss -0.243
Epoch 20 | Train Loss -0.286
Epoch 30 | Train Loss -0.352
Epoch 40 | Train Loss -0.363
Epoch 50 | Train Loss -0.432
Epoch 60 | Train Loss -0.506
Epoch 70 | Train Loss -0.477
Epoch 80 | Train Loss -0.610
Epoch 90 | Train Loss -0.550
Epoch 100 | Train Loss -0.650

<matplotlib.collections.PathCollection at 0x7fd32cc36790>

import torch

animation = []
num_samples = 1000
inputs = torch.randn(num_samples, 2)
for i, layer in enumerate(reversed(model.flow._layers)):
    if animation:
        for p in [0.75, 0.5, 0.25]:
            animation.append((i, animation[-1][1]*p + inputs.detach().numpy()*(1.0-p)))
    animation.append((i, inputs.detach().numpy()))
    inputs, _ = layer(inputs, inverse=True)

animation = [animation[0]]*10 + animation + [animation[-1]]*10
animation = animation + animation[::-1]

%matplotlib agg
import numpy as np
from IPython.display import Image
from matplotlib.animation import FuncAnimation

fig, ax = plt.subplots(figsize=(8,4))
plt.tight_layout()
ln = plt.scatter([], [])

def frame(i):
    layer, data = animation[i]
    ax.set_title(f"Flow Layer: {layer+1} of {len(model.flow._layers)}")
    ln.set_offsets(data)
    ax.set_xlim(np.min(data[:,0])-0.5, np.max(data[:,0])+0.5)
    ax.set_ylim(np.min(data[:,1])-0.5, np.max(data[:,1])+0.5)
    return ln,

fig.set_size_inches(8, 4, True)
ani = FuncAnimation(fig, frame, frames=list(range(len(animation))), interval=10, repeat=True, blit=False)
ani.save("2 Normalizing Flows.gif")
Image(url='2 Normalizing Flows.gif')

%matplotlib inline
plt.figure(figsize=(16,9))
before, after = animation[0][1][:100,:], animation[len(animation)//2][1][:100,:].copy()
after[:,0] += 5.0
for (x1, y1), (x2, y2) in zip(before, after):
    if np.random.random() > 0.1:
        plt.scatter([x1, x2], [y1, y2], marker = 'o', color=["r", "b"])
    else:
        plt.plot([x1, x2], [y1, y2], "--", marker = 'o')
#plt.scatter(before[:,0], before[:,1], color="b")
#plt.scatter(after[:,0], after[:,1], color="r")

echoflow_df = model.sample(num_samples=1000)

#----
from copulas.multivariate import GaussianMultivariate

copula = GaussianMultivariate()
copula.fit(df)
copula_df = copula.sample(1000)

#------
from ctgan import CTGANSynthesizer

ctgan = CTGANSynthesizer()
ctgan.fit(df)
ctgan_df = ctgan.sample(1000)

/Users/kevz/opt/anaconda3/envs/echoflow/lib/python3.8/site-packages/sklearn/utils/deprecation.py:143: FutureWarning: The sklearn.utils.testing module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.utils. Anything that cannot be imported from sklearn.utils is now part of the private API.
  warnings.warn(message, FutureWarning)
/Users/kevz/opt/anaconda3/envs/echoflow/lib/python3.8/site-packages/sklearn/utils/validation.py:67: FutureWarning: Pass n_components=10 as keyword args. From version 0.25 passing these as positional arguments will result in an error
  warnings.warn("Pass {} as keyword args. From version 0.25 "
/Users/kevz/opt/anaconda3/envs/echoflow/lib/python3.8/site-packages/sklearn/utils/validation.py:67: FutureWarning: Pass n_components=10 as keyword args. From version 0.25 passing these as positional arguments will result in an error
  warnings.warn("Pass {} as keyword args. From version 0.25 "

Epoch 1, Loss G: 0.0186, Loss D: 0.0461
Epoch 2, Loss G: -0.0088, Loss D: 0.0308
Epoch 3, Loss G: -0.0178, Loss D: 0.0465
Epoch 4, Loss G: -0.0112, Loss D: 0.0352
Epoch 5, Loss G: -0.0749, Loss D: 0.0501
Epoch 6, Loss G: -0.0434, Loss D: -0.0065
Epoch 7, Loss G: -0.0816, Loss D: 0.0059
Epoch 8, Loss G: -0.1145, Loss D: -0.0182
Epoch 9, Loss G: -0.1111, Loss D: -0.0369
Epoch 10, Loss G: -0.1490, Loss D: -0.0999
Epoch 11, Loss G: -0.1923, Loss D: -0.2093
Epoch 12, Loss G: -0.2507, Loss D: -0.1584
Epoch 13, Loss G: -0.2600, Loss D: -0.2598
Epoch 14, Loss G: -0.2905, Loss D: -0.2561
Epoch 15, Loss G: -0.3785, Loss D: -0.4029
Epoch 16, Loss G: -0.4014, Loss D: -0.5031
Epoch 17, Loss G: -0.4509, Loss D: -0.7288
Epoch 18, Loss G: -0.5261, Loss D: -0.7866
Epoch 19, Loss G: -0.6436, Loss D: -1.0426
Epoch 20, Loss G: -0.6144, Loss D: -1.2297
Epoch 21, Loss G: -0.7031, Loss D: -1.4010
Epoch 22, Loss G: -0.6613, Loss D: -1.5356
Epoch 23, Loss G: -0.7486, Loss D: -1.7074
Epoch 24, Loss G: -0.6783, Loss D: -1.8557
Epoch 25, Loss G: -0.7389, Loss D: -1.7605
Epoch 26, Loss G: -0.8345, Loss D: -1.8561
Epoch 27, Loss G: -0.8375, Loss D: -1.6751
Epoch 28, Loss G: -0.8852, Loss D: -1.5007
Epoch 29, Loss G: -0.9060, Loss D: -1.4957
Epoch 30, Loss G: -0.9102, Loss D: -1.2487
Epoch 31, Loss G: -1.0303, Loss D: -0.9371
Epoch 32, Loss G: -1.0190, Loss D: -0.8207
Epoch 33, Loss G: -1.0111, Loss D: -0.6011
Epoch 34, Loss G: -1.1496, Loss D: -0.4978
Epoch 35, Loss G: -1.2245, Loss D: -0.4232
Epoch 36, Loss G: -1.1854, Loss D: -0.3956
Epoch 37, Loss G: -1.2087, Loss D: -0.2473
Epoch 38, Loss G: -1.2302, Loss D: -0.2930
Epoch 39, Loss G: -1.2948, Loss D: -0.2669
Epoch 40, Loss G: -1.2661, Loss D: -0.3108
Epoch 41, Loss G: -1.2112, Loss D: -0.4249
Epoch 42, Loss G: -1.1491, Loss D: -0.5751
Epoch 43, Loss G: -1.2612, Loss D: -0.6512
Epoch 44, Loss G: -1.1156, Loss D: -0.8536
Epoch 45, Loss G: -1.1199, Loss D: -0.8210
Epoch 46, Loss G: -1.0663, Loss D: -1.0004
Epoch 47, Loss G: -1.1911, Loss D: -1.0190
Epoch 48, Loss G: -1.1268, Loss D: -1.2239
Epoch 49, Loss G: -1.2701, Loss D: -1.1039
Epoch 50, Loss G: -1.2590, Loss D: -0.9063
Epoch 51, Loss G: -1.2703, Loss D: -0.9106
Epoch 52, Loss G: -1.1734, Loss D: -0.7327
Epoch 53, Loss G: -1.1200, Loss D: -0.7415
Epoch 54, Loss G: -1.2540, Loss D: -0.6991
Epoch 55, Loss G: -1.1072, Loss D: -0.6870
Epoch 56, Loss G: -1.1942, Loss D: -0.7860
Epoch 57, Loss G: -1.1557, Loss D: -0.9378
Epoch 58, Loss G: -1.2369, Loss D: -0.7973
Epoch 59, Loss G: -1.2105, Loss D: -0.9420
Epoch 60, Loss G: -1.2373, Loss D: -1.0463
Epoch 61, Loss G: -1.1908, Loss D: -1.0182
Epoch 62, Loss G: -1.1133, Loss D: -1.0742
Epoch 63, Loss G: -1.1308, Loss D: -1.0755
Epoch 64, Loss G: -1.1058, Loss D: -1.0437
Epoch 65, Loss G: -1.0367, Loss D: -1.0882
Epoch 66, Loss G: -1.1599, Loss D: -1.0182
Epoch 67, Loss G: -0.9952, Loss D: -0.8387
Epoch 68, Loss G: -0.9895, Loss D: -0.8783
Epoch 69, Loss G: -1.0820, Loss D: -0.5924
Epoch 70, Loss G: -1.0012, Loss D: -0.7107
Epoch 71, Loss G: -0.9543, Loss D: -0.5602
Epoch 72, Loss G: -1.0246, Loss D: -0.4503
Epoch 73, Loss G: -0.9536, Loss D: -0.5506
Epoch 74, Loss G: -1.0457, Loss D: -0.3733
Epoch 75, Loss G: -1.0131, Loss D: -0.4134
Epoch 76, Loss G: -1.0671, Loss D: -0.5443
Epoch 77, Loss G: -1.1301, Loss D: -0.5082
Epoch 78, Loss G: -1.0976, Loss D: -0.6363
Epoch 79, Loss G: -1.0837, Loss D: -0.8867
Epoch 80, Loss G: -1.0284, Loss D: -1.0275
Epoch 81, Loss G: -1.0708, Loss D: -1.2676
Epoch 82, Loss G: -1.0427, Loss D: -1.2755
Epoch 83, Loss G: -0.9604, Loss D: -1.1389
Epoch 84, Loss G: -0.9721, Loss D: -1.3012
Epoch 85, Loss G: -0.9213, Loss D: -1.0286
Epoch 86, Loss G: -0.8522, Loss D: -0.8756
Epoch 87, Loss G: -0.7632, Loss D: -0.7728
Epoch 88, Loss G: -0.8879, Loss D: -0.4324
Epoch 89, Loss G: -0.9121, Loss D: -0.3359
Epoch 90, Loss G: -0.9421, Loss D: -0.1858
Epoch 91, Loss G: -0.8643, Loss D: -0.0971
Epoch 92, Loss G: -0.9237, Loss D: -0.0001
Epoch 93, Loss G: -0.8668, Loss D: 0.0448
Epoch 94, Loss G: -0.8508, Loss D: -0.0132
Epoch 95, Loss G: -0.8679, Loss D: -0.1154
Epoch 96, Loss G: -0.8538, Loss D: -0.3189
Epoch 97, Loss G: -0.8329, Loss D: -0.3361
Epoch 98, Loss G: -0.7238, Loss D: -0.4478
Epoch 99, Loss G: -0.7826, Loss D: -0.3892
Epoch 100, Loss G: -0.7111, Loss D: -0.2747
Epoch 101, Loss G: -0.7643, Loss D: -0.3702
Epoch 102, Loss G: -0.6601, Loss D: -0.2765
Epoch 103, Loss G: -0.6477, Loss D: -0.4186
Epoch 104, Loss G: -0.6715, Loss D: -0.3651
Epoch 105, Loss G: -0.5326, Loss D: -0.4281
Epoch 106, Loss G: -0.6217, Loss D: -0.5463
Epoch 107, Loss G: -0.5517, Loss D: -0.4937
Epoch 108, Loss G: -0.5091, Loss D: -0.4605
Epoch 109, Loss G: -0.4553, Loss D: -0.5801
Epoch 110, Loss G: -0.3853, Loss D: -0.6059
Epoch 111, Loss G: -0.3832, Loss D: -0.4931
Epoch 112, Loss G: -0.3960, Loss D: -0.4379
Epoch 113, Loss G: -0.3529, Loss D: -0.5796
Epoch 114, Loss G: -0.3796, Loss D: -0.5441
Epoch 115, Loss G: -0.3556, Loss D: -0.3659
Epoch 116, Loss G: -0.3073, Loss D: -0.5421
Epoch 117, Loss G: -0.2893, Loss D: -0.6636
Epoch 118, Loss G: -0.3211, Loss D: -0.5642
Epoch 119, Loss G: -0.2161, Loss D: -0.5835
Epoch 120, Loss G: -0.1647, Loss D: -0.5339
Epoch 121, Loss G: -0.1106, Loss D: -0.4994
Epoch 122, Loss G: -0.1449, Loss D: -0.3238
Epoch 123, Loss G: -0.0977, Loss D: -0.4864
Epoch 124, Loss G: -0.1453, Loss D: -0.4777
Epoch 125, Loss G: -0.1399, Loss D: -0.5016
Epoch 126, Loss G: -0.0171, Loss D: -0.5279
Epoch 127, Loss G: 0.0781, Loss D: -0.7221
Epoch 128, Loss G: 0.0976, Loss D: -0.7292
Epoch 129, Loss G: 0.1928, Loss D: -1.0534
Epoch 130, Loss G: 0.2554, Loss D: -1.1082
Epoch 131, Loss G: 0.3202, Loss D: -1.3533
Epoch 132, Loss G: 0.4023, Loss D: -1.1840
Epoch 133, Loss G: 0.3947, Loss D: -1.1548
Epoch 134, Loss G: 0.5020, Loss D: -1.2564
Epoch 135, Loss G: 0.6184, Loss D: -1.0490
Epoch 136, Loss G: 0.6725, Loss D: -0.9611
Epoch 137, Loss G: 0.7280, Loss D: -0.9741
Epoch 138, Loss G: 0.8168, Loss D: -0.9213
Epoch 139, Loss G: 0.8303, Loss D: -0.9053
Epoch 140, Loss G: 0.8566, Loss D: -0.9899
Epoch 141, Loss G: 0.9760, Loss D: -1.0215
Epoch 142, Loss G: 1.1160, Loss D: -1.2679
Epoch 143, Loss G: 1.1129, Loss D: -1.3706
Epoch 144, Loss G: 1.2057, Loss D: -1.4646
Epoch 145, Loss G: 1.1999, Loss D: -1.4563
Epoch 146, Loss G: 1.3380, Loss D: -1.4586
Epoch 147, Loss G: 1.2402, Loss D: -1.4347
Epoch 148, Loss G: 1.2568, Loss D: -1.4758
Epoch 149, Loss G: 1.3034, Loss D: -1.5865
Epoch 150, Loss G: 1.2778, Loss D: -1.6568
Epoch 151, Loss G: 1.3181, Loss D: -1.6223
Epoch 152, Loss G: 1.3955, Loss D: -1.7063
Epoch 153, Loss G: 1.3946, Loss D: -1.6753
Epoch 154, Loss G: 1.4669, Loss D: -1.7317
Epoch 155, Loss G: 1.4351, Loss D: -1.6430
Epoch 156, Loss G: 1.4505, Loss D: -1.8546
Epoch 157, Loss G: 1.5449, Loss D: -1.6044
Epoch 158, Loss G: 1.4576, Loss D: -1.5617
Epoch 159, Loss G: 1.5405, Loss D: -1.3070
Epoch 160, Loss G: 1.5457, Loss D: -1.6511
Epoch 161, Loss G: 1.7220, Loss D: -1.4746
Epoch 162, Loss G: 1.6910, Loss D: -1.5877
Epoch 163, Loss G: 1.8141, Loss D: -1.8678
Epoch 164, Loss G: 1.8353, Loss D: -1.6927
Epoch 165, Loss G: 1.7766, Loss D: -1.7656
Epoch 166, Loss G: 1.8900, Loss D: -1.8455
Epoch 167, Loss G: 1.7898, Loss D: -1.9633
Epoch 168, Loss G: 1.9036, Loss D: -1.7915
Epoch 169, Loss G: 1.8980, Loss D: -1.8210
Epoch 170, Loss G: 1.9618, Loss D: -1.9754
Epoch 171, Loss G: 1.9644, Loss D: -1.8309
Epoch 172, Loss G: 1.7909, Loss D: -2.1171
Epoch 173, Loss G: 1.8592, Loss D: -2.0299
Epoch 174, Loss G: 1.8749, Loss D: -2.3053
Epoch 175, Loss G: 1.7876, Loss D: -2.3782
Epoch 176, Loss G: 1.9045, Loss D: -2.4340
Epoch 177, Loss G: 1.7827, Loss D: -2.6373
Epoch 178, Loss G: 1.7055, Loss D: -2.5032
Epoch 179, Loss G: 1.7639, Loss D: -2.4069
Epoch 180, Loss G: 1.7336, Loss D: -2.4906
Epoch 181, Loss G: 1.6432, Loss D: -2.2937
Epoch 182, Loss G: 1.7854, Loss D: -2.1705
Epoch 183, Loss G: 1.7011, Loss D: -1.9169
Epoch 184, Loss G: 1.7280, Loss D: -1.9639
Epoch 185, Loss G: 1.8949, Loss D: -1.9455
Epoch 186, Loss G: 1.8718, Loss D: -2.0666
Epoch 187, Loss G: 1.9942, Loss D: -2.0417
Epoch 188, Loss G: 2.1457, Loss D: -2.1869
Epoch 189, Loss G: 2.0683, Loss D: -2.1986
Epoch 190, Loss G: 2.1045, Loss D: -2.2265
Epoch 191, Loss G: 2.1761, Loss D: -2.4311
Epoch 192, Loss G: 2.1594, Loss D: -2.4677
Epoch 193, Loss G: 2.1733, Loss D: -2.4785
Epoch 194, Loss G: 2.1443, Loss D: -2.4790
Epoch 195, Loss G: 2.0630, Loss D: -2.2714
Epoch 196, Loss G: 2.0614, Loss D: -2.3900
Epoch 197, Loss G: 2.0278, Loss D: -2.5413
Epoch 198, Loss G: 2.0790, Loss D: -2.4669
Epoch 199, Loss G: 1.9710, Loss D: -2.6489
Epoch 200, Loss G: 2.0048, Loss D: -2.2388
Epoch 201, Loss G: 2.0611, Loss D: -2.4497
Epoch 202, Loss G: 2.1429, Loss D: -2.5873
Epoch 203, Loss G: 2.0077, Loss D: -2.5879
Epoch 204, Loss G: 2.0901, Loss D: -2.6632
Epoch 205, Loss G: 2.0838, Loss D: -2.7802
Epoch 206, Loss G: 2.0216, Loss D: -2.7509
Epoch 207, Loss G: 2.0458, Loss D: -2.6516
Epoch 208, Loss G: 2.1515, Loss D: -2.5217
Epoch 209, Loss G: 2.0505, Loss D: -2.8524
Epoch 210, Loss G: 2.1325, Loss D: -2.6729
Epoch 211, Loss G: 2.1394, Loss D: -2.4977
Epoch 212, Loss G: 2.0353, Loss D: -2.6402
Epoch 213, Loss G: 2.2289, Loss D: -2.7521
Epoch 214, Loss G: 2.1723, Loss D: -2.8613
Epoch 215, Loss G: 2.1450, Loss D: -2.8816
Epoch 216, Loss G: 2.2011, Loss D: -2.8288
Epoch 217, Loss G: 2.2493, Loss D: -3.3624
Epoch 218, Loss G: 2.0373, Loss D: -2.9492
Epoch 219, Loss G: 2.1170, Loss D: -3.1125
Epoch 220, Loss G: 2.0623, Loss D: -2.9412
Epoch 221, Loss G: 2.0564, Loss D: -2.8479
Epoch 222, Loss G: 2.0038, Loss D: -2.6160
Epoch 223, Loss G: 2.0325, Loss D: -2.5881
Epoch 224, Loss G: 2.0703, Loss D: -2.7975
Epoch 225, Loss G: 2.0520, Loss D: -2.7406
Epoch 226, Loss G: 2.1087, Loss D: -2.6677
Epoch 227, Loss G: 2.0558, Loss D: -2.8955
Epoch 228, Loss G: 2.0997, Loss D: -2.8635
Epoch 229, Loss G: 2.0423, Loss D: -2.8859
Epoch 230, Loss G: 2.1814, Loss D: -2.7471
Epoch 231, Loss G: 2.1695, Loss D: -2.9966
Epoch 232, Loss G: 2.2432, Loss D: -3.1865
Epoch 233, Loss G: 2.2534, Loss D: -3.1304
Epoch 234, Loss G: 2.2750, Loss D: -3.0837
Epoch 235, Loss G: 2.1427, Loss D: -3.0589
Epoch 236, Loss G: 2.1377, Loss D: -2.9294
Epoch 237, Loss G: 2.1839, Loss D: -3.0754
Epoch 238, Loss G: 2.1048, Loss D: -3.0462
Epoch 239, Loss G: 2.0367, Loss D: -3.0427
Epoch 240, Loss G: 2.0954, Loss D: -3.1331
Epoch 241, Loss G: 2.0329, Loss D: -2.9186
Epoch 242, Loss G: 2.1275, Loss D: -3.1956
Epoch 243, Loss G: 2.1732, Loss D: -3.1239
Epoch 244, Loss G: 2.1581, Loss D: -3.2789
Epoch 245, Loss G: 2.1447, Loss D: -3.2709
Epoch 246, Loss G: 2.2186, Loss D: -3.0476
Epoch 247, Loss G: 2.1449, Loss D: -3.2314
Epoch 248, Loss G: 2.1429, Loss D: -3.0765
Epoch 249, Loss G: 2.2048, Loss D: -3.5900
Epoch 250, Loss G: 2.1555, Loss D: -3.4545
Epoch 251, Loss G: 2.1801, Loss D: -3.4485
Epoch 252, Loss G: 2.2512, Loss D: -3.4843
Epoch 253, Loss G: 2.2078, Loss D: -3.6149
Epoch 254, Loss G: 2.2443, Loss D: -3.5558
Epoch 255, Loss G: 2.1366, Loss D: -3.2953
Epoch 256, Loss G: 2.1251, Loss D: -3.3531
Epoch 257, Loss G: 2.0502, Loss D: -3.3458
Epoch 258, Loss G: 2.1738, Loss D: -3.3758
Epoch 259, Loss G: 2.0999, Loss D: -3.5996
Epoch 260, Loss G: 2.1857, Loss D: -3.6469
Epoch 261, Loss G: 2.2199, Loss D: -3.5684
Epoch 262, Loss G: 2.2074, Loss D: -3.7502
Epoch 263, Loss G: 2.1945, Loss D: -3.6650
Epoch 264, Loss G: 2.2254, Loss D: -3.9411
Epoch 265, Loss G: 2.2120, Loss D: -3.6680
Epoch 266, Loss G: 2.3173, Loss D: -3.8879
Epoch 267, Loss G: 2.2653, Loss D: -3.9813
Epoch 268, Loss G: 2.2608, Loss D: -3.4992
Epoch 269, Loss G: 2.2234, Loss D: -3.7144
Epoch 270, Loss G: 2.2439, Loss D: -3.4729
Epoch 271, Loss G: 2.1858, Loss D: -3.5375
Epoch 272, Loss G: 2.2002, Loss D: -3.8484
Epoch 273, Loss G: 2.2694, Loss D: -3.5786
Epoch 274, Loss G: 2.2606, Loss D: -3.6222
Epoch 275, Loss G: 2.2411, Loss D: -3.7043
Epoch 276, Loss G: 2.2248, Loss D: -3.8401
Epoch 277, Loss G: 2.1606, Loss D: -3.9935
Epoch 278, Loss G: 2.2319, Loss D: -4.1673
Epoch 279, Loss G: 2.2833, Loss D: -4.2293
Epoch 280, Loss G: 2.3112, Loss D: -4.2375
Epoch 281, Loss G: 2.1604, Loss D: -4.0040
Epoch 282, Loss G: 2.2728, Loss D: -4.1549
Epoch 283, Loss G: 2.2237, Loss D: -4.1131
Epoch 284, Loss G: 2.2349, Loss D: -3.8488
Epoch 285, Loss G: 2.2760, Loss D: -3.9987
Epoch 286, Loss G: 2.2157, Loss D: -4.0173
Epoch 287, Loss G: 2.2435, Loss D: -4.0976
Epoch 288, Loss G: 2.3455, Loss D: -4.4764
Epoch 289, Loss G: 2.3667, Loss D: -3.8304
Epoch 290, Loss G: 2.3954, Loss D: -4.2998
Epoch 291, Loss G: 2.3236, Loss D: -4.1624
Epoch 292, Loss G: 2.2916, Loss D: -4.1199
Epoch 293, Loss G: 2.3128, Loss D: -4.1032
Epoch 294, Loss G: 2.3832, Loss D: -4.4515
Epoch 295, Loss G: 2.2991, Loss D: -4.4757
Epoch 296, Loss G: 2.3276, Loss D: -4.3763
Epoch 297, Loss G: 2.3185, Loss D: -4.3259
Epoch 298, Loss G: 2.3393, Loss D: -4.2809
Epoch 299, Loss G: 2.2827, Loss D: -4.1953
Epoch 300, Loss G: 2.3268, Loss D: -4.3975

fig, axes = plt.subplots(nrows=2, ncols=2, figsize=(12, 8))
axes[0,0].scatter(df["x"], df["y"], color="#e77c8d")
axes[0,0].set_title("Real")
axes[1,0].scatter(ctgan_df["x"], ctgan_df["y"], color="#56ad74")
axes[1,0].set_title("CTGAN")
axes[0,1].scatter(copula_df["x"], copula_df["y"], color="#5ea5c5")
axes[0,1].set_title("Copulas")
axes[1,1].scatter(echoflow_df["x"], echoflow_df["y"], color="#4878d0")
axes[1,1].set_title("EchoFlow")
fig.tight_layout()