Posts by Collection

activities

portfolio

publications

High Quality Depth Estimation from Monocular Images Based on Depth Prediction and Enhancement Sub-Networks

IEEE International Conference on Multimedia and Expo (ICME), 2018

(ChatGPT-Generated) The paper proposes a novel depth estimation framework based on deep convolutional neural network (CNN) to learn the mapping between monocular images and depth maps, which includes a depth prediction network based on the ResNet architecture to infer the scene depth from color image and a depth enhancement network concatenated to the end of the depth prediction network to obtain a high resolution depth map, and experimental results show that the proposed method outperforms other methods on benchmark RGB-D datasets and achieves state-of-the-art performance.

Recommended citation: X. Duan, X. Ye, Y. Li and H. Li, "High Quality Depth Estimation from Monocular Images Based on Depth Prediction and Enhancement Sub-Networks," 2018 IEEE International Conference on Multimedia and Expo (ICME), San Diego, CA, USA, 2018, pp. 1-6, doi: 10.1109/ICME.2018.8486539. https://ieeexplore.ieee.org/document/8486539

Retrospective Thinking based Multi-Agent System for Wireless Video Transmissions

 IEEE International Conference on Communications, 2021

(ChatGPT-Generated) The paper proposes a retrospective thinking based multi-agent (ReTMA) system for wireless video transmissions, which includes a retrospective agent on the feedback loop and facilitate the complex training issue of the multiple agents, and experimental results show that the proposed method outperforms other plain feedback models in terms of similarity index measure (SSIM), peak signal-to-noise ratio (PSNR), and classification accuracy at different mask ratios.

Recommended citation: Y. Li et al., "Retrospective Thinking based Multi-Agent System for Wireless Video Transmissions," ICC 2021 - IEEE International Conference on Communications, Montreal, QC, Canada, 2021, pp. 1-6, doi: 10.1109/ICC42927.2021.9500274. https://ieeexplore.ieee.org/document/9500274

LIGS: Learnable Intrinsic-Reward Generation Selection for Multi-Agent Learning

ICLR , 2022

(ChatGPT-Generated) The paper introduces a new framework called LIGS, which improves coordination and performance of multi-agent reinforcement learners by introducing a learnable intrinsic-reward generation selection algorithm. LIGS enables systems of multi-agent reinforcement learners to quickly solve environments with sparse rewards and demonstrates superior performance in challenging tasks in Foraging and StarCraft II.

Recommended citation: David Henry Mguni, Taher Jafferjee, Jianhong Wang, Oliver Slumbers, Nicolas Perez-Nieves, Feifei Tong, Li Yang, Jiangcheng Zhu, Yaodong Yang, and Jun Wang. "Ligs: Learnable intrinsic-reward generation selection for multi-agent learning." arXiv preprint arXiv:2112.02618 (2021). http://academicpages.github.io/files/paper2.pdf

MAFENN: Multi-Agent Feedback Enabled Neural Network for Wireless Channel Equalization

IEEE Global Communications Conference (GLOBECOM), 2022

(ChatGPT-Generated) The paper proposes a Multi-Agent Feedback Enabled Neural Network (MAFENN) equalizer for wireless channel equalization, which consists of a specific learnable feedback agent and two feed-forward agents that improve the ability to remove wireless inter-symbol interference (ISI) in receiving ends. The MAFENN framework is formulated into a three-player Stackelberg Game, which helps to optimize and train the model more efficiently, and experimental results show that the proposed method outperforms other equalizers in terms of symbol error rate (SER) performance at different signal-to-noise ratio (SNR) settings for both linear and nonlinear channels.

Recommended citation: Y. Li et al., "MAFENN: Multi-Agent Feedback Enabled Neural Network for Wireless Channel Equalization," 2021 IEEE Global Communications Conference (GLOBECOM), Madrid, Spain, 2021, pp. 1-6, doi: 10.1109/GLOBECOM46510.2021.9685522. https://ieeexplore.ieee.org/document/9685522

Multi-Agent Feedback Enabled Neural Networks for Intelligent Communications

IEEE Transactions on Wireless Communications, 2022

(ChatGPT-Generated) The paper proposes a novel multi-agent feedback enabled neural network (MAFENN) framework for intelligent communications, which consists of three fully cooperative intelligent agents that have stronger feedback learning capabilities and more intelligence on feature abstraction, denoising or generation. The MAFENN framework is theoretically formulated into a three-player Feedback Stackelberg game and is shown to outperform traditional or DL-based equalizers in wireless fading channels with inter-symbol interference (ISI), demonstrating its effectiveness and robustness in complex channel environments.

Recommended citation: Fanglei Sun, Yang Li, Ying Wen, Jingchen Hu, Jun Wang, Yang Yang, and Kai Li. "Multi-Agent Feedback Enabled Neural Networks for Intelligent Communications." IEEE Transactions on Wireless Communications 21, no. 8 (2022): 6167-6179. https://ieeexplore.ieee.org/abstract/document/9705656

Cross-Utterance Conditioned VAE for Non-Autoregressive Text-to-Speech

ACL, 2022

(ChatGPT-Generated) The paper proposes a cross-utterance conditioned VAE for non-autoregressive text-to-speech, which allows the prosody features generated by the TTS system to be related to the context, and demonstrates superior performance in terms of naturalness, intelligibility, and prosody diversity on LJ-Speech and LibriTTS data. The proposed CUC-VAE estimates a posterior probability distribution of the latent prosody features for each phoneme by conditioning on acoustic features, speaker information, and text features obtained from both past and future sentences.

Recommended citation: Yang Li, Cheng Yu, Guangzhi Sun, Hua Jiang, Fanglei Sun, Weiqin Zu, Ying Wen, Yang Yang, and Jun Wang. "Cross-Utterance Conditioned VAE for Non-Autoregressive Text-to-Speech." arXiv preprint arXiv:2205.04120 (2022). 1(3). https://arxiv.org/abs/2205.04120

Cooperative Open-ended Learning Framework for Zero-shot Coordination

ICML, 2023

(ChatGPT-Generated) Zero-shot coordination in cooperative AI remains a significant challenge, with previous approaches often leading to cooperative incompatibility. The Cooperative Open-ended LEarning (COLE) framework addresses this issue by using graph theory to assess and identify cooperative abilities, outperforming current methods in the Overcooked game environment.

Recommended citation: Yang Li, Shao Zhang, Jichen Sun, Yali Du, Ying Wen, Xinbing Wang, and Wei Pan. "Cooperative Open-ended Learning Framework for Zero-shot Coordination." arXiv preprint arXiv:2302.04831 (2023). https://arxiv.org/abs/2302.04831

JiangJun: Mastering Xiangqi by Tackling Non-Transitivity in Two-Player Zero-Sum Games

Transactions on Machine Learning Research (TMLR), 2023

(ChatGPT-Generated) The paper investigates the non-transitive nature of the Chinese board game Xiangqi and introduces the JiangJun algorithm, blending Monte-Carlo Tree Search (MCTS) with Policy Space Response Oracles (PSRO) to approximate a Nash equilibrium. Tested through a WeChat mini program, the algorithm boasts a remarkable 99.41% win rate against human players, confirming its effectiveness in handling non-transitivity.

Recommended citation: Yang Li, Kun Xiong, Yingping Zhang, Jiangcheng Zhu, Stephen Marcus McAleer, Wei Pan, Jun Wang, Zonghong Dai, & Yaodong Yang (2023). JiangJun: Mastering Xiangqi by Tackling Non-Transitivity in Two-Player Zero-Sum Games. Transactions on Machine Learning Research. https://openreview.net/forum?id=MMsyqXIJuk

reading

LLM & Human-Machine Collaboration Reading Party: Cycle 0

Published:

Introduction

Before the summer holiday, we successfully organized a month-long reading party focused on LLM and robotics (see Cycle 0: LLM and robotics). During this month, we passionately explored various applications of LLM in the field of robotics. From the Sept. 2023, we hope to revive the reading party. Our theme will continue to revolve around LLM and human-machine collaboration. To make our discussions more in-depth and specific, we will set a particular research direction each month. Each direction will be hosted by a PhD student studying the area, whose primary task will be to recommend and propose relevant outstanding papers. Subsequently, participating students will delve deep into these papers, sharing insights and engaging in discussions. Note: It’s acceptable if the paper doesn’t utilize LLM, as long as it falls under the topic and possesses potential for LLM-enhanced improvements.

LLM & Human-Machine Collaboration Reading Party

Published:

Introduction

Before the summer holiday, we successfully organized a month-long reading party focused on LLM and robotics (see Cycle 0: LLM and robotics). During this month, we passionately explored various applications of LLM in the field of robotics. From the Sept. 2023, we hope to revive the reading party. Our theme will continue to revolve around LLM and human-machine collaboration. To make our discussions more in-depth and specific, we will set a particular research direction each month. Each direction will be hosted by a PhD student studying the area, whose primary task will be to recommend and propose relevant outstanding papers. Subsequently, participating students will delve deep into these papers, sharing insights and engaging in discussions. Note: It’s acceptable if the paper doesn’t utilize LLM, as long as it falls under the topic and possesses potential for LLM-enhanced improvements.

talks

teaching

ICML 2023 Volunteer at Hawaii

activities, ICML, 2023

Aloha! Volunteering at ICML 2023 was a privilege, allowing me to discuss with leading machine learning experts from around the globe.