simular

machine learning + handwritten digits
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
 i 
 i 
0-24
0-43
0-44
0-15
0-28
0-19
0-37
0-29
0-13
0-17
0-48
0-49
0-47
0-16
0-22
0-25
0-27
0-35
0-33
0-32
0-36
0-45
0-30
0-18
0-7
0-1
0-10
0-8
0-23
0-46
0-31
0-9
0-41
0-2
0-4
0-3
0-38
0-21
0-6
0-34
0-11
0-50
0-5
0-20
0-14
0-39
0-12
0-40
0-26
0-42
1-37
1-48
1-21
1-14
1-8
1-7
1-6
1-26
1-50
1-11
1-43
1-32
1-19
1-33
1-39
1-13
1-3
1-20
1-18
1-30
1-34
1-22
1-12
1-17
1-25
1-49
1-47
1-5
1-4
1-24
1-44
1-16
1-10
1-40
1-1
1-35
1-46
1-38
1-36
1-28
1-9
1-29
1-41
1-15
1-27
1-42
1-2
1-23
1-45
1-31
2-34
2-30
2-17
2-23
2-19
2-44
2-48
2-46
2-27
2-13
2-30
2-49
2-20
2-39
2-14
2-47
2-25
2-43
2-3
2-9
2-15
2-26
2-10
2-31
2-33
2-16
2-22
2-11
2-2
2-28
2-32
2-21
2-4
2-50
2-41
2-42
2-7
2-37
2-8
2-29
2-1
2-12
2-38
2-6
2-45
2-18
2-5
2-24
2-36
2-35
3-11
3-5
3-46
3-40
3-48
3-26
3-30
3-17
3-3
3-29
3-1
3-16
3-45
3-50
3-25
3-14
3-20
3-27
3-18
3-37
3-22
3-41
3-24
3-39
3-13
3-43
3-2
3-42
3-47
3-6
3-10
3-9
3-44
3-32
3-36
3-12
3-49
3-23
3-4
3-21
3-7
3-15
3-31
3-33
3-34
3-35
3-19
3-8
3-38
3-28
4-3
4-34
4-11
4-47
4-15
4-49
4-18
4-19
4-16
4-12
4-14
4-9
4-26
4-21
4-23
4-42
4-2
4-4
4-33
4-38
4-20
4-41
4-35
4-37
4-29
4-27
4-45
4-48
4-30
4-17
4-36
4-1
4-13
4-46
4-25
4-10
4-32
4-7
4-39
4-44
4-31
4-8
4-50
4-24
4-43
4-22
4-28
4-5
4-40
4-6
5-40
5-38
5-36
5-11
5-15
5-16
5-4
5-43
5-17
5-29
5-34
5-8
5-47
5-41
5-32
5-26
5-14
5-30
5-25
5-44
5-23
5-9
5-27
5-3
5-2
5-18
5-33
5-12
5-5
5-1
5-24
5-28
5-49
5-20
5-6
5-45
5-39
5-22
5-13
5-48
5-31
5-19
5-7
5-46
5-21
5-10
5-50
5-42
5-37
5-35
6-31
6-7
6-28
6-20
6-22
6-26
6-4
6-6
6-11
6-13
6-9
6-44
6-37
6-34
6-5
6-48
6-43
6-32
6-3
6-33
6-24
6-49
6-38
6-25
6-39
6-8
6-47
6-12
6-1
6-23
6-15
6-45
6-17
6-40
6-2
6-29
6-30
6-19
6-50
6-35
6-42
6-16
6-14
6-36
6-21
6-46
6-27
6-41
6-18
6-10
7-16
7-10
7-13
7-8
7-17
7-35
7-14
7-30
7-3
7-32
7-20
7-18
7-31
7-21
7-48
7-40
7-29
7-45
7-42
7-49
7-28
7-2
7-43
7-25
7-7
7-11
7-9
7-4
7-24
7-33
7-34
7-6
7-47
7-15
7-44
7-26
7-19
7-36
7-39
7-41
7-50
7-12
7-46
7-23
7-37
7-1
7-5
7-27
7-38
7-22
8-17
8-36
8-31
8-20
8-8
8-32
8-44
8-46
8-10
8-48
8-37
8-6
8-38
8-27
8-28
8-26
8-43
8-35
8-12
8-19
8-24
8-41
8-11
8-29
8-9
8-5
8-1
8-25
8-21
8-4
8-15
8-30
8-3
8-47
8-18
8-42
8-16
8-49
8-23
8-50
8-22
8-34
8-7
8-13
8-39
8-2
8-45
8-14
8-40
8-33
9-7
9-26
9-39
9-1
9-45
9-22
9-10
9-14
9-8
9-44
9-32
9-16
9-3
9-41
9-43
9-46
9-13
9-2
9-19
9-23
9-17
9-40
9-38
9-48
9-27
9-49
9-9
9-24
9-4
9-5
9-50
9-18
9-28
9-11
9-34
9-42
9-37
9-35
9-25
9-47
9-21
9-6
9-31
9-20
9-12
9-33
9-15
9-36
9-30
9-29

Generative adversarial networks (GANs) are an architecture for training two models simultaneously – the generator (artist) learns to create images that look real and the discriminator (art critic) learns to apart real images from fakes. The conditional generative adversarial network (CGAN) is a type of GAN that involves the conditional on a class label or data from other modalities of images.

This experiment with a CGAN generates images for each digit. The website shows a series of images from 50 epochs. The model was trained with THE MNIST DATABASE using Python and Keras/TensorFlow in Jupyter Notebook.



REFERENCES


THE MNIST DATABASE of handwritten digits
Generative Adversarial Networks
Deep Convolutional Generative Adversarial Network



© 2021 by Claudia Wasem