<p>Anyone using Go for production Machine Learning? Most advice online advocates using Python but Go as a language seems a better fit for compute-heavy tasks. Does Python just have better ML tools?</p>
<hr/>**评论：**<br/><br/>tinrab: <pre><p>Training and using trained models are different tasks. As far as I know, it's easier to train models in Python than in other languages. But when it comes to serving them in production, any language will do. Models like decision trees or neural nets would be very easy to implement in Go. For example, TensorFlow has Go bindings which can load pre-trained models.</p></pre>everdev: <pre><p>It seems like Go would be faster at training models than Python. Is it just a lack of a quality ML training package for Go that people choose Python?</p></pre>seankhliao: <pre><p>Python is the primary choice for data analysis/machine learning beside it is easy and it can be used in a scripting manner which works really well in an interactive session (think jupyter/ipython notebooks) for exploratory work</p></pre>sbinet: <pre><p>Go does have some ways to be used from Jupyter: <a href="https://github.com/gopherdata/gophernotes" rel="nofollow">https://github.com/gopherdata/gophernotes</a></p></pre>seankhliao: <pre><p>Most of the heavy machine learning frameworks are written in C and have python bindings which are what we commonly use, so no, not much of a performance gain to be had</p></pre>
