Diffusion Autoencoders: Toward a Meaningful and Decodable Representation
- Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, Supasorn Suwajanakorn
- Computer ScienceComputer Vision and Pattern Recognition
- 30 November 2021
This paper explores the possibility of using DPMs for representation learning and seeks to extract a meaningful and decodable representation of an input image via autoencoding and can encode any image into a two-part latent code, allowing near-exact reconstruction.
Improved image classification explainability with high-accuracy heatmaps
- Konpat Preechakul, S. Sriswasdi, B. Kijsirikul, E. Chuangsuwanich
- Computer ScienceiScience
- 15 February 2022
CProp: Adaptive Learning Rate Scaling from Past Gradient Conformity
- Konpat Preechakul, B. Kijsirikul
- Computer ScienceArXiv
- 24 December 2019
CProp is proposed, a gradient scaling method, which acts as a second-level learning rate adapting throughout the training process based on cues from past gradient conformity, which could apply to any existing optimizer extending its learning rate scheduling capability.
High resolution weakly supervised localization architectures for medical images
- Konpat Preechakul, S. Sriswasdi, B. Kijsirikul, E. Chuangsuwanich
- Computer ScienceArXiv
- 22 October 2020
This work proposes Pyramid Localization Network (PYLON), a model for high-accuracy weakly-supervised localization that achieved 0.62 average point localization accuracy on NIH's Chest X-Ray 14 dataset, compared to 0.45 for a traditional CAM model.
Set Prediction in the Latent Space
- Konpat Preechakul, Chawan Piansaddhayanon, Burin Naowarat, Tirasan Khandhawit, S. Sriswasdi, E. Chuangsuwanich
- Computer ScienceNeural Information Processing Systems
- 2021
A method for learning the distance function by performing the matching in the latent space learned from encoding networks to provide sufficient conditions for permutation stability which begets an algorithm to improve the overall model convergence.