Data-centric AI: From ImageNet to Open-Source FinRL and FinGPT
Xiao-Yang Liu
The deep learning revolution began with the winning of the ImageNet 2012 challenge by a convolutional neural network AlexNet, and then PyTorch, an open-source framework, has facilitated model innovations, further advancing the field. Recently, data has become increasingly important in AI, leading to the emergence of data-centric AI. In this talk, I will share my experience working with the ImageNet-21K dataset, which is a much larger version of the static ImageNet-1K dataset, and how I leveraged the interplay between machine learning, signal processing, and computing to achieve further progress.
As the creator of open-source projects FinRL, ElegantRL, and FinGPT, I will discuss my experiences using these tools to tackle the challenging domain of the financial market. Financial reinforcement learning presents unique challenges due to the highly dynamic nature of financial data. FinRL and FinGPT build upon the success of deep learning, and I will focus mainly on an automatic data curation pipeline that addresses the challenges of handling financial data.
This seminar will be held both online & in person. You are welcome to join us either in South Hall or via Zoom.
Xiao-Yang Liu is a Ph.D. candidate in the Department of Electrical Engineering at Columbia University. He received his M.S. in electrical engineering from Columbia University in 2018 and his Ph.D. in the Department of Computer Science and Engineering at Shanghai Jiao Tong University in 2017. His research interests include tensor theory and high-performance tensor computation, deep learning, big data analysis and privacy, and open-source AI4Finance.
Xiao-Yang has authored two textbooks: one on tensors for data processing and another on reinforcement learning for cyber-physical systems. He serves as a PC member for various conferences, including NeurIPS, ICML, ICLR, KDD, AAAI, ACM ICAIF, and MM, and has chaired sessions at IJCAI 2019. Additionally, he has organized multiple workshops, including the NeurIPS 2020/2021 First/Second Workshop on Quantum Tensor Networks in Machine Learning, IJCAI 2020 Workshop on Tensor Networks Representations in Machine Learning, and NeurIPS 2019/2020 Workshop on Machine Learning for Autonomous Driving.