Computation / Visualization Framework
We introduce a new kind of computing environment. Our central innovation is to encode data, meta-data and machine-learned models in a minimalist structure that a human can see and manipulate in visual space. We will demonstrate how this tool is applied to predict rare, severe, adverse events in patients, such as those suffering from Acute Renal Failure, a condition affecting tens of millions of Americans. We will also describe some general benefits to the user, including querying of the database without writing computer code and simplifying the data conditioning steps required for machine learning. These innovations facilitate rapid, iterative refinement of machine learning models, while introducing new challenges, such as the need to guard against overfitting by the human user. We will solicit feedback on the design principles for this system of connecting human and machine, and its potential for facilitating collaborative computing. Helynx is a computing platform created by an MD Ph.D., a physicist, and two students/former-students in the Caltech CNS program. Our largest client is a Fortune 500 integrated healthcare provider.