AKOOL Research

Authors: Yaxuan Zhu, Jianwen Xie, Ying Nian Wu, Ruiqi Gao
Conference: The Twelfth International Conference on Learning Representations (ICLR), 2024
‍
This research presents an innovative method for learning energy-based models using cooperative diffusion recovery likelihood. Our approach leverages the strengths of cooperative learning and diffusion processes to improve the training efficiency and effectiveness of energy-based models.
‍
Read the Paper

Authors: Weinan Song, Yaxuan Zhu, Lei He, Ying Nian Wu, Jianwen Xie
Archive: ArXiv, 2024
‍
We propose a progressive energy-based cooperative learning framework for multi-domain image-to-image translation. This method addresses the challenges of domain adaptation and translation by progressively refining the learning process, resulting in superior performance across multiple domains.

Read the Paper

Authors: Deqian Kong, Dehong Xu, Minglu Zhao, Bo Pang, Jianwen Xie, Andrew Lizarraga, Yuhao Huang, Sirui Xie, Ying Nian Wu

In this paper, we introduce the Latent Plan Transformer, a novel framework that treats planning as latent variable inference. Our approach combines the strengths of transformers and latent variable models to achieve robust and efficient planning in complex environments.

‍Conference: The 38th Conference on Neural Information Processing Systems (NeurIPS 2024)

Read the Paper

Authors: Khoa Doan, Jianwen Xie, Yaxuan Zhu, Yang Zhao, Ping Li

CoopHash introduces a novel approach to supervised image hashing through cooperative learning. By leveraging Variational Markov Chain Monte Carlo (MCMC) teaching, it simultaneously optimizes a multipurpose descriptor and a contrastive pair generator. This innovative method enhances the efficiency and accuracy of image hashing, making it a significant advancement in the field.

Read the Paper

Authors: Peiyu Yu, Dinghuai Zhang, Hengzhi He, Xiaojian Ma, Ruiyao Miao, Yifan Lu, Yasi Zhang, Deqian Kong, Ruiqi Gao, Jianwen Xie, Guang Cheng, Ying Nian Wu

This paper introduces a novel black-box optimization method known as the Latent Energy-Based Odyssey. The approach focuses on expanded exploration within an energy-based latent space, enhancing the optimization process. By leveraging the latent space’s energy landscape, the method improves the efficiency and effectiveness of optimization tasks, making it a significant advancement in the field of machine learning and optimization.

Read the Paper

Authors: Deqian Kong, Yuhao Huang, Jianwen Xie, Edouardo Honig, Ming Xu, Shuanghong Xue, Pei Lin, Sanping Zhou, Sheng Zhong, Nanning Zheng, Ying Nian Wu

We propose the Latent Prompt Transformer (LPT), a novel generative model for challenging problem of molecule design. Experiments demonstrate that LPT not only effectively discovers useful molecules across single-objective, multi-objective, and structure-constrained optimization tasks, but also exhibits strong sample efficiency.

Read the Paper