How Alibaba’s Experts Tackle Large‑Scale Matching and Deep Learning Challenges

On June 29, 2017, Alibaba’s Xixi Campus in Hangzhou hosted the inaugural Data Mining Frontier Development and Future Forum, jointly organized by Alibaba Group, the Chinese Society of Computational Linguistics, and KDD China. Nearly 300 experts from leading universities and companies attended.

Key speakers included Alibaba iDST head Jin Rong, Ant Financial AI Department Technical Director Li Xiaolong, NTU professor Lin Zhiren (IEEE/ACM/AAAI Fellow), Tsinghua associate professor Cui Peng, and CAS researcher Luo Ping.

Jin Rong: Large‑Scale Online Precise Matching in Taobao “Ask Everyone”

Precise matching aims to assign tasks to the most suitable actors, optimizing overall reward parameters. The “Ask Everyone” feature matches user questions with potential answerers at massive scale, leveraging Alibaba’s abundant data on both askers and responders to improve task allocation.

Lin Zhiren: Importance of Single‑Machine and Distributed Settings for Big‑Data Machine Learning

In the era of big data, datasets often exceed the capacity of a single machine, making distributed learning essential. However, single‑machine setups remain valuable for sampling and rapid prototyping. The choice depends on specific workflow constraints.

Panel Discussion: Machine Learning, Data Mining, and Deep Learning

Panelists (Alibaba senior algorithm experts, Ant Financial AI director, Tsinghua and CAS researchers, Zhejiang University lecturer) explored definitions, relationships, and trends. They noted that deep learning now dominates machine learning research, offering standardized optimization methods, modular model components, and strong performance across domains.

Challenges highlighted included deep learning’s lack of interpretability, high data requirements, robustness issues, and the need for theoretical foundations such as hierarchical Bayesian models.

Industry‑academia collaboration was emphasized: industrial data and computing resources can accelerate research, while academic curiosity drives innovative problem formulation beyond immediate commercial goals.