Quantum Generative Adversarial Networks to Reproduce Calorimeters Outputs
- Slides: 12
Quantum Generative Adversarial Networks to Reproduce Calorimeters Outputs 19 May 2021 QGAN to Reproduce Calorimeters Outputs Su Yeon Chang, Sofia 1 Vallecorsa
Continuous Variable (CV) Quantum Computing Gate Unitary Displacement Rotation Gaussian Squeezing Beamsplitter Kerr Non-gaussian https: //strawberryfields. readthedocs. io/en/stable/introduction. html 19 May 2021 QGAN to Reproduce Calorimeters Outputs 2
CV Quantum Neural Networks Interferometer Squeezing Displacement Non-gaussian Orthogonal Diagonal https: //doi. org/10. 1038/ncomms 13795 19 May 2021 QGAN to Reproduce Calorimeters Outputs 3
CV Quantum GAN Generator https: //arxiv. org/pdf/1811. 04968. pdf Discriminator Update parameters using Quantum Gradient descent https: //arxiv. org/pdf/1909. 07806. pdf https: //arxiv. org/pdf/1612. 01789. pdf 19 May 2021 QGAN to Reproduce Calorimeters Outputs 4
Implementation 19 May 2021 QGAN to Reproduce Calorimeters Outputs 5
Simple classifier test ü Reduce the problem size into 1 dimension and bin into 3 pixels → 3 qumodes ü Use CV Neural Network to classify real and fake data (Case depth = 3) Actual Real (1) Fake (0) Real (1) 100% 5% Fake (0) 0% 95% Predicted 19 May 2021 QGAN to Reproduce Calorimeters Outputs 6
Rescale sigmoid function 19 May 2021 QGAN to Reproduce Calorimeters Outputs 7
CV GAN test ü Latent vector of size 1, following normal distribution ü Using Adam Optimizer depthd = 3, depthg = 4 19 May 2021 depthd = 3, depthg = 6 QGAN to Reproduce Calorimeters Outputs 8
CV GAN test depthd = 3, depthg = 6 19 May 2021 QGAN to Reproduce Calorimeters Outputs 9
Increase batchsize depthd = 3, depthg = 6 19 May 2021 QGAN to Reproduce Calorimeters Outputs 10
Current step ü Testing GAN with different parameters ü Test different loss function (Wasserstein Loss) ü Difficulty : Requires a long time for simulation Ex : 50 min for one epoch with 1, 000 samples, depthd = 3, depthg = 6 ü Come back to classical network to test it is possible to reproduce the targeted “image” set 19 May 2021 QGAN to Reproduce Calorimeters Outputs 11
Questions? THANK YOU 19 May 2021 QGAN to Reproduce Calorimeters Outputs 12
- Quantum generative adversarial learning
- Generative adversarial networks
- Melody randford
- Spectral normalization gan
- Generative adversarial network
- Quantum physics vs quantum mechanics
- Classical physics
- A link layer protocol for quantum networks
- Basestore iptv
- Difference between virtual and datagram circuit
- Voice conversion
- Hexiangnan
- Adversarial patch