Data Mining
Data Mining
I use data mining methods to learn from weak labels, sparse evidence, operational logs, and field experiment settings.
Learning from imperfect evidence
How can learning systems extract reliable signals when labels, behavior, and operational data are incomplete or noisy?
Scope
Weak supervision, robust learning, topic modeling, continual graph learning, and digital business experiments form the modern method line.
Weak and robust learning
Methods for learning when labels are scarce, noisy, or created through weak supervision.
Text and graph mining
Topic modeling, continual graph convolution, keyword graph learning, and data augmentation.
Digital business experiments
Live-streaming analytics and AI assistance studied through field experiment settings.
Papers in this area
Data Mining
Neural Topic Modeling Based on Cycle Adversarial Training and Contrastive Learning
Continual Graph Convolutional Network for Text Classification
EPiDA: An Easy Plug-in Data Augmentation Framework for High Performance Text Classification
DP-SSL: Towards Robust Semi-Supervised Learning with a Few Labeled Samples
Weakly-Supervised Text Classification Based on Keyword Graph
Effectiveness of AI Assistance in Live-Streaming: A Randomized Field Experiment
The Sales Data Sells: Effects of Real-Time Sales Analytics on Live Streaming Selling
Related patents
Patents