Revision History for Blending Two Styles: Generating...

Accepted Edit by TMLR

• 26 Dec 2024, 00:16 Coordinated Universal Time

• Venue: Accepted by TMLR
• Venueid: TMLR
• Bibtex:

  @article{
  macdonald2024blending,
  title={Blending Two Styles: Generating Inter-domain Images with Middle{GAN}},
  author={Collin MacDonald and Zhendong Chu and John Stankovic and Huajie Shao and Gang Zhou and Ashley Gao},
  journal={Transactions on Machine Learning Research},
  issn={2835-8856},
  year={2024},
  url={https://openreview.net/forum?id=t7vWCHmwbG},
  note={}
  }

• Authors: (empty)(readers deleted)
• Authorids: (empty)(readers deleted)
• Supplementary Material: (empty)(readers deleted)

• Note – Pdate: 26 Dec 2024, 00:16 Coordinated Universal Time
• Note – Writers: TMLR

Camera Ready Revision Edit by Authors

• 23 Dec 2024, 18:18 Coordinated Universal Time

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• Authors: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• Authorids: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• PDF: pdf
• Changes Since Last Submission:

  We revised our manuscript based on the Reviewer's thoughtful critiques. We have included what we changed in our individual responses to the Reviewers.

Camera Ready Revision Edit by Authors

• 23 Dec 2024, 15:26 Coordinated Universal Time

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• Authors: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• Authorids: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• PDF: pdf
• Changes Since Last Submission:

  We revised our manuscript based on the Reviewer's thoughtful critiques. We have included what we changed in our individual responses to the Reviewers.

Camera Ready Revision Edit by Authors

• 23 Dec 2024, 15:18 Coordinated Universal Time

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• Authors: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• Authorids: Collin MacDonald, Zhendong Chu, John Stankovic, Huajie Shao, Gang Zhou, Ashley Gao
• PDF: pdf
• Changes Since Last Submission:

  We revised our manuscript based on the Reviewer's thoughtful critiques. We have included what we changed in our individual responses to the Reviewers.

Edit Edit by TMLR

• 22 Dec 2024, 01:07 Coordinated Universal Time

• Venue: Decision pending for TMLR
• Venueid: TMLR/Decision_Pending

Revision Edit by Authors

• 23 Oct 2024, 17:03 Coordinated Universal Time

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• PDF: pdf
• Submission Length: Long submission (more than 12 pages of main content)
• Changes Since Last Submission:

  We revised our manuscript based on the Reviewer's thoughtful critiques. We have included what we changed in our individual responses to the Reviewers.

Revision Edit by Authors

• 22 Oct 2024, 16:44 Coordinated Universal Time

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• PDF: pdf
• Submission Length: Long submission (more than 12 pages of main content)
• Changes Since Last Submission:

  We revised our manuscript based on the Reviewer's thoughtful critiques. We have included what we changed in our individual responses to the Reviewers.

Under Review Edit by TMLR

• 28 Aug 2024, 00:43 Coordinated Universal Time

• Bibtex:

  @article{
  anonymous2024blending,
  title={Blending Two Styles: Generating Inter-domain Images with Middle{GAN}},
  author={Anonymous},
  journal={Submitted to Transactions on Machine Learning Research},
  year={2024},
  url={https://openreview.net/forum?id=t7vWCHmwbG},
  note={Under review}
  }

• Venue: Under review for TMLR
• Venueid: TMLR/Under_Review

• Note – Odate: 28 Aug 2024, 00:43 Coordinated Universal Time
• Note – Readers: everyone

Edit Edit by TMLR

• 02 Aug 2024, 02:32 Coordinated Universal Time (modified: 28 Aug 2024, 00:43 Coordinated Universal Time)

• Assigned Action Editor: Mingming Gong
• Venueid: TMLR/Assigned_AE
• Venue: TMLR Assigned AE

Edit Edit by TMLR

• 26 Jul 2024, 20:03 Coordinated Universal Time (modified: 28 Aug 2024, 00:43 Coordinated Universal Time)

• Venue: TMLR Assigning AE
• Venueid: TMLR/Assigning_AE

Revision Edit by Authors

• 21 Jul 2024, 19:26 Coordinated Universal Time (modified: 28 Aug 2024, 00:43 Coordinated Universal Time)

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• PDF: pdf
• Submission Length: Long submission (more than 12 pages of main content)
• Changes Since Last Submission:

  We updated Section 3 (Theory) to fix some typos and added a remark to summarize the proof and restate its purpose. We also updated the layout of the Figures to make them appear naturally embedded in the text.

  Nothing else has been changed.

Submission Edit by TMLR

• 20 Jul 2024, 18:10 Coordinated Universal Time (modified: 28 Aug 2024, 00:43 Coordinated Universal Time)

• Title: Blending Two Styles: Generating Inter-domain Images with MiddleGAN
• Abstract: From celebrity faces to cats and dogs, humans enjoy pushing the boundaries of art by blending existing concepts together in new ways. With the rise of generative artificial intelligence, machines are increasingly capable of creating new images. Generative Adversarial Networks (GANs) generate images similar to their training data but struggle to blend images from distinct datasets. This paper introduces MiddleGAN, a novel GAN variant that blends inter-domain images from two distinct input sets. By incorporating a second discriminator, MiddleGAN forces the generator to create images that fool both discriminators, thus capturing the qualities of both input sets. We also introduce a blend ratio hyperparameter to control the weighting of the input sets and compensate for datasets of different complexities. Evaluating MiddleGAN on the CelebA dataset, we demonstrate that it successfully generates images that lie between the distributions of the input sets, both mathematically and visually. An additional experiment verifies the viability of MiddleGAN on handwritten digit datasets (DIDA and MNIST). We provide a proof of optimal convergence for the neural networks in our architecture and show that MiddleGAN functions across various resolutions and blend ratios. We conclude with potential future research directions for MiddleGAN.
• PDF: pdf
• Submission Length: Long submission (more than 12 pages of main content)
• Venue: Submitted to TMLR
• Venueid: TMLR/Submitted

• Note – License: CC BY-SA 4.0
• Note – Signatures: TMLR/Paper3039/Authors
• Note – Readers: TMLR, TMLR/Paper3039/Action_Editors, TMLR/Paper3039/Authors
• Note – Writers: TMLR, TMLR/Paper3039/Authors