Meta-Learning and Auto-Labeling for Machine Learning

A non-transitory computer-readable medium having instructions stored thereon that are executable by a computing device to perform operations comprising:

• determining, based on output of a benchmark machine learning model for a set of testing examples, a performance threshold for machine learning models;
• comparing output of a meta-learning machine learning model for a new set of examples with known labels for examples in the new set of examples, wherein the new set of examples includes one or more new features;
• determining, based on the comparing, whether a current performance of the meta-learning model satisfies the performance threshold;
• and
• in response to determining that the current performance of the meta-learning model does not satisfy the performance threshold, automatically triggering retraining of the meta-learning model.

A method, comprising:

• performing, by a computer system, an auto-labeling process for a set of examples used to iteratively train a machine learning model, wherein the auto-labeling process includes:
• determining whether known labels for examples in the set of examples are correct, wherein the determining includes comparing classifications output by the machine learning model for examples in the set of examples and the known labels for the set of examples with label error information;
• and
• assigning, based on determining that one or more of the known labels are incorrect, penalties to examples corresponding to the one or more incorrect labels.