Zero-Knowledge Discovery Using Data Smashing
Abstract
Project ZeD addresses zero knowledge inference in sequential data streams: the task of finding-models from raw data when we do not necessarily know the correct model structure a priori. Absence of such prior knowledge is becoming increasingly common for the complex questions we are now asking in biology, social systems, physics and engineering. We cannot reduce this exercise to one of simply tuning parameters and/or model calibration. Sparsity of reported de no-vo modeling paradigms that allow automated abduction of good models from raw data is a key bottleneck in automated problem solving. Our effort is designed to address this emerging gap. Leveraging fundamentally new insights into automated inference, such as the principle of data smashing and abductive learning of generative models of quantized stochastic processes, we en-vision transformative breakthroughs in automated problem solving.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2021
- Accession Number
- AD1152831
Entities
Organizations
- University of Chicago