throw_away_777

12 ms·

61.
by throw_away_777 10y ago
It is actually much easier to apply a random forest (or really gradient boosted decision tree, which almost strictly dominates random forests) than a linear regression. Decision tree methods require far less data preprocessing than linear r
62.
by throw_away_777 10y ago
When you have a reputation for great culture and good pay, and are one of the best known tech companies, you get enough applicants that you can use pretty much whatever hiring process you want.
63.
by throw_away_777 10y ago
The fact that squared error is differentiable is not irrelevant. You can solve some machine learning models faster with differentiable objectives (most notably xgboost). Speed is important, you need to optimize your models and the longer it
64.
by throw_away_777 10y ago
This is the difference between practice and theory. In theory differential objectives don't matter, in practice for medium to large datasets they make machine learning a lot faster. Speed is critical, as you need to be able to iterate
65.
by throw_away_777 10y ago
There is a Kaggle competition right now that uses mean absolute error, and this makes the problem substantially harder. For a practical discussion of techniques used to solve machine learning problems that use mae see the forums in: https:
66.
by throw_away_777 10y ago
Often times getting close to an answer that is approximately correct is better than trying to find the perfect solution. Most real world problems, especially in analyzing data, don't have perfect answers. For example approximations are
67.
by throw_away_777 10y ago
If your analysis involves data, you can't skip the statistics. If there is no analysis of data than go ahead and skip the stats all you want. You need to try and determine the uncertainty in your results, both systematic and statistica
68.
by throw_away_777 10y ago
The main advantage of the frequentist approach is that you can do the calculations much more easily. Bayesian statistics is great, but often the calculations are much more difficult because of your distribution of priors. You can make up si
69.
by throw_away_777 10y ago
There are different types of colliders, leptonic and hadronic colliders. Hadrons, usually protons, are composite particles so when you smash them together a wide range of output is produced. You can absolutely see unexpected things, and a l
70.
by throw_away_777 10y ago
The democratic party really needs to start prioritizing the needs of working class Americans. Too many policies clearly benefit corporations and the rich, at the expense of workers. Obama has been a huge disappointment, he talked of hope an
71.
by throw_away_777 10y ago
There is substantial evidence that a lot of people voted for Trump because they did not like Clinton. If Clinton was a popular candidate I don't see a path to Trump winning, but she was a historically unpopular candidate, only Trump ha
72.
by throw_away_777 10y ago
By the time you have a PhD a lot of problems may seem simple, especially in your field, but you spent 6 years of study most people haven't. The truth is that machine learning is the intersection of statistics, data analysis and softwar
73.
by throw_away_777 10y ago
Well then those people can refuse the pay and still take the interview.
74.
by throw_away_777 10y ago
No one is proposing sending all the immigrants back to where they came from, and if they did the economy would not cease to function. Maybe part of the problem is that companies like Google hire teams of 90% immigrants while Americans can&#
75.
by throw_away_777 10y ago
Democracy means having to go with the majority. And Trump isn't Hitler, that is just ridiculous.
76.
by throw_away_777 10y ago
I agree that manufacturing jobs are now automated, so they aren't coming back. But your other points are not true - imagine the US halted all imports. If you halt all imports, companies either produce in the US or don't sell to US
77.
by throw_away_777 10y ago
I don't really see the difference, but it should be noted that given Clinton's minority support there weren't many minorities supporting third party candidates, and there was a significant third party candidate vote this year
78.
by throw_away_777 10y ago
This is also what 1 in 6 looks like, or what 1 in 10 looks like, or what 1 in 100 looks like. You can't definitively judge probabilistic predictions based on just 1 event.
79.
by throw_away_777 10y ago
Clinton lost because her support among white voters was weak, not because of minority voters.
80.
by throw_away_777 10y ago
Sometimes there are no definitive answers. The world we live in is not deterministic, at least with our current knowledge of the world.
81.
by throw_away_777 10y ago
Tariffs can absolutely change the fact that your citizens are too expensive. You can argue that the consequences are not worth tariffs, but you can't say that they don't promote production in your country.
82.
by throw_away_777 10y ago
Government in the US has been extremely dysfunctional for a while now, and Clinton is the embodiment of the establishment politician. This may explain why so many people were willing to vote for an outsider candidate, despite his obvious fl
83.
by throw_away_777 10y ago
The candidate gets no benefit from doing your take home test. In an interview the candidate learns something about the company. Also the ~1h requirement just isn't realistic if you give an open-ended test - you penalize the people who
84.
by throw_away_777 10y ago
This product already exists, check out https://www.isidewith.com/
85.
by throw_away_777 10y ago
A lot of people's HN accounts, at least for the people who post a lot, are not anonymous. Alternatively if you post a lot it is possible to reconstruct who you are. If comments were being down voted you wouldn't see them at the to
86.
by throw_away_777 10y ago
I don't understand what you are trying to say - political correctness is all about preventing people from saying things that aren't "politically" correct. It is a tool for preventing discussion about topics which might o
87.
by throw_away_777 10y ago
I looked through your history but couldn't find a comment where you were being not politically correct, but I didn't try all that hard. Regardless there are certain sentiments that you can't express even though they may be tr
88.
by throw_away_777 10y ago
Throwaway accounts allow people to express views that aren't politically correct. If all your views are politically correct this may seem like a bad thing, but for people with views that aren't politically correct throw away accou
89.
by throw_away_777 10y ago
I think point 7) needs work. Often times people use words like interpretable to avoid having to think about the data - usually in the context of linear or logistic regression. The model seems "interpretable" because the coefficien
90.
by throw_away_777 10y ago
I've always disliked the term data cleaning for the reasons you mention - it doesn't tell me anything about what is meant by "cleaning".

More ›