Does there have to be a tradeoff between speed and accuracy? Using rapid prototyping, you can have both.
In this talk, the Jaxon team discusses how to minimize your model’s time-to-production while maximizing its usefulness and accuracy.
They discuss how to solve real-world business problems, starting with the concept of the “problem spec” and how it can be used to frame problems in terms that ML models are equipped to solve. Next, they dive into model creation and iteration, plus some common stumbling blocks and techniques to improve models. Finally, they discuss learning rates and multi-stage model training.
