Open Internet by MindsNet

Achieving Sub-Millisecond Tabular Inference

The challenge is to develop ultrafast and tiny decision-making models for tabular classification that can run on commodity CPUs, competing with or beating gradient-boosted decision trees (GBDTs). Current methods have limitations in terms of inference speed and model size. There is a need for novel approaches that can provide sub-millisecond inference times.

Computing & Technology, Computer Science, Machine Learning