请问现在真的是机器学习比较好吗

pcheng11 发表于 2019-3-4 16:18
机器学习入坑需谨慎，我就是跳坑了，楼主要思考一下自己未来的发展方向和自己的兴趣。

what do you mean?
a good ML engineer still commands the highest pay level... higher than pure SWE, and surely higher than DS.
有的公司MLE有不同的pay ladder

ML product engineer - use well known algorithms, leverage relatively mature ML infra, and build ML into product, mostly working on feature engineering, could train using mature methods, "use" inference (instead of optimizing inference deep into the stack), track results maybe via AB test. Output is measured by lift in product metrics. Highest expertise is not in deep knowledge in modeling, but in software engineering - "can ship sth to production" is the most important.

ML infra engineer - build ML infra, does not focus on modeling, but should be deeper in the stack, more like engineer than scientists. Most likely not even remotely like scientists. often works with distributed systems, has deep knowledge about data infra, sometimes hardware (if deals with inference on the edge). Can help other teams work better is the most important.

ML scientist - work on new algo (new algo), would read papers, often write papers, output is measured not by publication but often by some product related metrics. Need a solid foot in product domain, and another foot in fairly deep ML algo. Many of them work on search ranking, recommender systems etc. Can build prototypes, and can help engineers develop models that move product in the right direction is the most important.

research scientist - cutting edge research, output is measured by publication. Most important: advancing the field.

机器学习是个大方向吧

好像也不是一个转行的好方向？？？

jackalex 发表于 2019/02/16 05:28:52
我在搞ML。PHD已经毕业了在公司做研发。 主要工作是看paper，推公式，实现，然后看看能不能用到公司的project上还有就是想办法做分布式计算，加快速度。基本不调参。调参花不了多少时间
业余时...

楼主为啥还刷leetcode啊？工作需要？跳槽需要？

pcheng11 发表于 2019-3-5 08:18
机器学习入坑需谨慎，我就是跳坑了，楼主要思考一下自己未来的发展方向和自己的兴趣。

具体啥坑...