Our research mission is to bridge the gap between theory and practice in machine learning for problems that arise in financial markets and social network analysis. Our lab aims to develop robust forecasting models for financial instruments under harsh conditions (e.g., excessively large number of features and notoriously low signal-to-noise), and to develop techniques to tackle ultra-sparse graph-learning problems in social network analysis (e.g., high-dimensional, low signal-to-noise, and combinatorial structure).
Our research is primarily funded by NSF: NSF SMALL, NSF CRII Award 2018, NSF OAC Award 2018, Activision/Blizzard gift and Rutherford Fellowship at the Alan Turing Institute (UK). We also thank William & Mary for the start-up grant.
- Theory. We design statistically sound and computationally efficient methods optimized for financial market and social network analysis problems, generally by marrying random graph and high-dim statistics techniques.
- Practice. We examine power of limitations of non-linear models (including but not limited to deep learning and tree-based techniques), generally by integrating domain-specific knowledge with architectural design and developing engineering techniques to circumvent overfitting.
- Open source software. We build open source tools to support massive large-scale ML compute loads. We disseminate our new tools and ensure that others can replicate our experiments.
Theory (learning with combinatorial structure)
- [Finance] Qiong Wu, Felix Ming Fai Wong, Zhenming Liu, Yanhua Li, and Varun Kanade, “Adaptive reduced rank regression” [PDF]
- [Finance] W. Yun, Z. Liu, V. Kanade, C. Wang, “High-dimensional block signal recovery”
- [Social network] Ao Liu, Qiong Wu, Zhenming Liu and Lirong Xia, “Near-Neighbor Methods in Random Preference Completion” [PDF]
- [Social network] Cheng Li, Felix Wong, Zhenming Liu, Varun Kanade, “From which world is your graph”[PDF]
- [Finance] Qiong Wu, Zheng Zhang, Andrea Pizzoferrato, Mihai Cucuringu, and Zhenming Liu, “End-to-end financial instrument”[PDF]
- [Social network] Felix Ming Fai Wong, Zhenming Liu, Mung Chiang, “On the efficiency of social recommender systems”
We are currently developing systems that co-optimize between ML algorithms and SVD solvers; our work is supported by an NSF award. More details will be available soon.
Other research activities
We also work with system researchers to design systems with performance guarantees, such as:
- Authors DistCache: Provable Load Balancing for Large-Scale Storage Systems with Distributed Caching
- Authors Improving user QoE for residential broadband: Adaptive traffic management at the network edge
- Authors Optimizing the one big switch abstraction in software-defined networks
We provide ML solutions to industrial practitioners. We have extensive experience across different sectors, including pharmaceutical (see our recent work), gaming (Activision-Blizzard), shared riding services (see our work [PDF]), and FinTech. Please email me for inquiries.
We do not take new projects related to generating alpha signals at this time.