Similarity based learning machine and methods of similarity based machine learning
OwnerYang Changinventor
Life of this patent
Estimating the term…
Abstract
In accordance with aspects and embodiments, an improved similarity based learning machine and methods of similarity based machine learning are provided. More specifically, the learning machines and machine learning methods of the present disclosure advantageously define subjects by attributes, assign a first similarity score to each of the subjects, from the first similarity score, calculate attribute scaling factors, and use the attribute scaling factors to generate an improved similarity score. In accordance with aspects and embodiments, the improved similarity scores may be used to improve machine learning.
Claims17
A system for machine learning, the system comprising:
- a computer;
- a non-transitory computer-readable medium having software stored therein, when executed by said computer, performs a machine learning method comprising the steps of:
- generating a similarix from received training data of N subjects, where each subject is defined by a small set of attributes, wherein the small set of attributes is less than or equal to three hundred attributes;
- assigning an initial similarity score to each of the subjects;
- calculating a scaling factor for each attribute;
- generating improved similarity scores using the scaling factors;
- using the improved similarity scores to predict future outcomes for new subjects;
- and
- wherein the future outcomes are predicted without overfitting to actual, observed outcomes.
A method of supervised machine learning, the method comprising the steps of:
- providing training data to a learning machine;
- generating initial similarity scores;
- generating improved similarity scores;
- using the improved similarity scores to model outcome;
- using the modeled outcome to predict future outcomes;
- wherein the training data comprises N subjects defined by k attributes;
- wherein the improved similarity score is calculated by first calculating an attribute scaling factor from the initial similarity scores;
- wherein the k attributes are less than or equal to three hundred attributes;
- and
- wherein the future outcomes are predicted without overfitting to actual, observed outcomes.
A method for treating a disease in a current patient, the method comprising the steps of:
- providing training data of prior patients in a clinical trial to a learning machine, the training data including for each prior patient attributes and patient outcome,
- generating initial similarity scores between prior patients;
- generating improved similarity scores between prior patients;
- using the improved similarity scores to model patient outcome;
- using the modeled outcome to predict an outcome of a current patient
- using the predicted outcome to arrive at a course of treatment for the current patient and
- wherein generating improved similarity scores between prior patients comprises calculating attribute scaling factors from the initial similarity scores and using the calculated attribute scaling factors to generate the improved similarity scores.
Dependencies are inferred from the claim text.