pidgan 0.1.3

GAN-based models to flash-simulate the LHCb PID detectors

PIDGAN

GAN-based models to flash-simulate the LHCb PID detectors

Generative Adversarial Networks

Algorithms*	Avail	Test	Lipschitzianity**	Design inspired by	Tutorial
GAN	✅	✅	❌	1, 8, 9
BceGAN	✅	✅	❌	2, 8, 9
LSGAN	✅	✅	❌	3, 8, 9
WGAN	✅	✅	✅	4, 9
WGAN_GP	✅	✅	✅	5, 9
CramerGAN	✅	✅	✅	6, 9
WGAN_ALP	✅	✅	✅	7, 9
BceGAN_GP	✅	✅	✅	2, 5, 9
BceGAN_ALP	✅	✅	✅	2, 7, 9

*each GAN algorithm is designed to operate taking conditions as input [10]

**the GAN training is regularized to ensure that the discriminator encodes a 1-Lipschitz function

Generators

Players	Avail	Test	Inherit from	Design inspired by
Generator	✅	✅	tf.keras.Model	1, 10
ResGenerator	✅	✅	Generator	1, 10, 11

Discriminators

Players	Avail	Test	Inherit from	Design inspired by
Discriminator	✅	✅	tf.keras.Model	1, 9, 10
ResDiscriminator	✅	✅	Discriminator	1, 9, 10, 11
AuxDiscriminator	✅	✅	ResDiscriminator	1, 9, 10, 11, 12

Other players

Players	Avail	Test	Inherit from
Classifier	✅	✅	Discriminator
ResClassifier	✅	✅	ResDiscriminator
AuxClassifier	✅	✅	AuxDiscriminator
MultiClassifier	✅	✅	Discriminator
MultiResClassifier	✅	✅	ResDiscriminator
AuxMultiClassifier	✅	✅	AuxDiscriminator

References

1. I.J. Goodfellow et al., "Generative Adversarial Networks", arXiv:1406.2661
2. A. Radford, L. Metz, S. Chintala, "Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks", arXiv:1511.06434
3. X. Mao et al., "Least Squares Generative Adversarial Networks", arXiv:1611.04076
4. M. Arjovsky, S. Chintala, L. Bottou, "Wasserstein GAN", arXiv:1701.07875
5. I. Gulrajani et al., "Improved Training of Wasserstein GANs", arXiv:1704.00028
6. M.G. Bellemare et al., "The Cramer Distance as a Solution to Biased Wasserstein Gradients", arXiv:1705.10743
7. D. Terjék, "Adversarial Lipschitz Regularization", arXiv:1907.05681
8. M. Arjovsky, L. Bottou, "Towards Principled Methods for Training Generative Adversarial Networks", arXiv:1701.04862
9. T. Salimans et al., "Improved Techniques for Training GANs", arXiv:1606.03498
10. M. Mirza, S. Osindero, "Conditional Generative Adversarial Nets", arXiv:1411.1784
11. K. He et al., "Deep Residual Learning for Image Recognition", arXiv:1512.03385
12. A. Rogachev, F. Ratnikov, "GAN with an Auxiliary Regressor for the Fast Simulation of the Electromagnetic Calorimeter Response", arXiv:2207.06329

Credits

Most of the GAN algorithms are an evolution of what provided by the mbarbetti/tf-gen-models repository. The BceGAN model is freely inspired by the TensorFlow tutorial Deep Convolutional Generative Adversarial Network and the Keras tutorial Conditional GAN. The WGAN_ALP model is an adaptation of what provided by the dterjek/adversarial_lipschitz_regularization repository.