wdroz

19 ms·

181.
by wdroz 9y ago
I like the absence of java-based tools. But IMO the Word count example is really verbose : https://github.com/WallarooLabs/wallaroo/blob/0.1.2/examples... With Apache Spark, the Word count example is a l
182.
by wdroz 9y ago
Posting your project on HN can help, but I think you already know that.
183.
by wdroz 9y ago
I buy stocks from near companies. I keep contacts from people in R&D. Some companies do partnership on research with universities. So you can easily network and hear about their public projects. But be very careful if you intent to crea
184.
by wdroz 9y ago
You also created the backend abstraction that let you customize a lot without real "raw TensorFlow". Thank you Mr. Chollet.
185.
by wdroz 9y ago
Python - Introspection, metaprogramming
186.
by wdroz 9y ago
I specialize 10-30% of my CV for the target role. I have a CV for data scientist and another for software engineer. I don't list the same weighted skills for both roles. In both cases, I list and describe relevant projects for each wor
187.
by wdroz 9y ago
"This device is not yet supported" more like Deeplearn_chrome_only.js
188.
by wdroz 9y ago
The problem is the GIL : https://wiki.python.org/moin/GlobalInterpreterLock Multiprocessing allows you to use all cores, but you can't share memory like in threads. So you have to Serialize/Deserialize a lot
189.
by wdroz 9y ago
Cool, so you can enhance your mental capacity by doing this test again and again. In less than a year I will be a genius!
190.
by wdroz 9y ago
Xonsh let you interact with other bash commands. >>> out = $(echo @(x + ' ' + y)) >>> out 'xonsh party\n' >>> @("ech" + "o") "hey" hey In raw python
191.
by wdroz 9y ago
I use Xonsh[0] for scripting now. I explain why in my blog[1], basically because we are in 2017 and I like python. [0]: http://xon.sh/ [1]: https://william-droz.com/xonsh-a-modern-shell-that-enable-py...
192.
by wdroz 9y ago
You can do a "regular" programming job and seeking business cases at your company where data science could help. After, meet your boss and tell him something like "I can make this process 10-20% faster with a 3 month projects
193.
by wdroz 9y ago
>>Getting too many false negatives is a very sure way of getting a lot of false positives. Why ? IMO if you reduce the False Acceptance Rate, you will up the False Reject Rate. Good chart at https://www.tractica.com/bi
194.
by wdroz 10y ago
Do you get extra rewards if you score well with random forests ?
195.
by wdroz 10y ago
Python 2.7 only is a no-go for me.
196.
by wdroz 10y ago
I will try scikit-plot, I usually waste time on plotting "by hand". I hope it works with Keras using wrappers.
197.
Xonsh – A modern shell that enable python in your terminal(william-droz.com)
2 points by wdroz 10y ago|0 comments
198.
by wdroz 10y ago
Experienced in Python? This will take you less than one minute.
199.
by wdroz 10y ago
When you see this post from your website, it's not recursive. Why ?
200.
by wdroz 10y ago
Kivy isn't well know but I use this often : https://kivy.org/