News for hackers. Stuff that matters.
Re:The Sigmoid Function in Logistic Regression by tacos on Tuesday May 17, 2016 @02:41PM (#11716158)
If the point of Hacker News is to be a sewing circle where we slowly reveal first principles of introductory textbooks, after someone makes the top of the homepage with a fundamentally wacky and incorrect analysis -- just so you can enjoy the comments -- then everything you propose makes perfect sense. Same process also works great in kindergarten classrooms. But at some point you have to stop rewarding people for drawing a yellow sun in crayon with a happy face on it or they never grow up.

Last week's idiotic neural net post involved data compression. It's now a top Google result atop research papers from the people who actually figured it out 30 years ago. I'm eager to see next week's thriller... see ya in the comments. If I get in early enough I'll fix it with love; if I get in late again I may or may not take the time to question why we keep doing things this way. And if I do a half-assed job with either, I'll certainly appreciate people telling me so. Might not work for everyone, but seems to open doors for me--including driver's side on the van.

Re:The Sigmoid Function in Logistic Regression by dang on Tuesday May 17, 2016 @04:05PM (#11716919)
Please stop making the thread worse. Other people already pointed out problems with the article; there was no need to pile on by calling names—which the HN guidelines ask you not to do—let alone take the thread so tediously off topic.

Even assuming you're 100% right, hauling in barrels of garbage and dumping them in the commons is not a good reaction.