Daily Productive Sharing 492 - Solve Important Problems

Daily Productive Sharing 492 - Solve Important Problems
Photo by stefan moertl / Unsplash

One helpful tip per day:)

Ben Kuhn looks back on his past and realizes that since the beginning of his studies, he has tended to solve hard problems. Such a choice certainly works in school because it gets you higher grades, but not necessarily in the real world. So he suggests that in reality, we need to focus on important problems, not hard problems.

This is very similar to training machine learning models. In school, we only need to consider the metrics of the model itself when we publish a paper, and of course the higher the better. However, in industry it's completely different, because the model metrics and the business needs are not always aligned, and the metrics that measure how good the model is are not just the metrics of the model itself, but there may be some external metrics, such as processing speed and so on. So in industry, training a model requires a wider range of considerations, and choosing how to measure a model is more important than choosing what model to use.

If you enjoy today's sharing, why not subscribe?

Need a superb CV, please try our CV Consultation


#reading

Ben Kuhn 回顾自己的过去,发现自读书开始,他就倾向于解决难题。这样的选择在学校里当然行得通,因为可以让你拿到更高的分数,但是在现实世界里却并不一定。所以他建议,在现实中,我们要着眼于重要问题,而不是难题。

这和训练机器学习模型很相似,在学校里,我们发论文时只需要考虑模型本身的指标,当然是越高越好;但是在工业界就完全不一样了,因为模型指标和商业需求并一定完全一致,而且衡量模型好坏的指标也不仅仅是模型本身的指标,可能还有一些外在的指标,比如处理速度等等。所以在工业界训练一个模型需要考虑的范围更广,选择如何衡量一个模型比选择什么模型更重要。

如果我们的内容对你有价值,不如付费支持我们 :)

需要更棒的简历,不妨试试我们的 CV Consultation

如果你也想成为更高效的人,欢迎加入我们的 TG group