We are a not-for-profit organization that focuses on addressing machine learning limitations using machine learning. [ML]² is a space for machine learning enthusiasts to collaborate and find solutions for such limitations.
The most practical problem with Deep Learning is the percieved lack of interpretability and explainability. We, at [ML]², are developing a framework to make every Deep Learning Model ineterpretable.
Make "interpretability" a first class method on a model object
Can the producer and consumer of an ML model be separated? IMLY does it by providing a reference dnn implementation to a target specificaton.
Provide every ML algorithm with an equivalent DNN implementation
ML is progressing at a brisk rate of 100 papers per month, modestly speaking. Curate that knowledge, and make it avaiable to both humans and algorithms.
Make ML machine-understandable.