Читать книгу Machine Learning For Dummies - John Paul Mueller, John Mueller Paul, Luca Massaron - Страница 18

Huge datasets require huge amounts of memory. Unfortunately, the requirements don’t end there. When you have huge amounts of data and memory, you must also have processors with multiple cores and high speeds. One of the problems that scientists are striving to solve is how to use existing hardware more efficiently. In some cases, waiting for days to obtain a result to a machine learning problem simply isn’t possible. The scientists who want to know the answer need it quickly, even if the result isn’t quite right. With this in mind, investments in better hardware also require investments in better science. This book considers some of the following issues as part of making your machine learning experience better:

 Obtaining a useful result: As you work through the book, you discover that you need to obtain a useful result first, before you can refine it. In addition, sometimes tuning an algorithm goes too far and the result becomes quite fragile (and possibly useless outside a specific dataset).

 Asking the right question: Many people get frustrated in trying to obtain an answer from machine learning because they keep tuning their algorithm without asking a different question. To use hardware efficiently, sometimes you must step back and review the question you’re asking. The question might be wrong, which means that even the best hardware will never find the answer.

 Relying on intuition too heavily: All machine learning questions begin as a hypothesis. A scientist uses intuition to create a starting point for discovering the answer to a question. Failure is more common than success when working through a machine learning experience. Your intuition adds the art to the machine learning experience, but sometimes intuition is wrong and you have to revisit your assumptions.

When you begin to realize the importance of environment to machine learning, you can also begin to understand the need for the right hardware and in the right balance to obtain a desired result. The current state-of-the-art systems actually rely on Graphical Processing Units (GPUs) to perform machine learning tasks. Relying on GPUs does speed the machine learning process considerably. A full discussion of using GPUs is outside the scope of this book, but you can read more about the topic at https://devblogs.nvidia.com/parallelforall/bidmach-machine-learning-limit-gpus/ and https://towardsdatascience.com/what-is-a-gpu-and-do-you-need-one-in-deep-learning-718b9597aa0d.