skierscott

10 ms·

61.
by skierscott 10y ago
This is an excellent post and I love the interactive demos. > gradient descent ... ball rolling down a hill Gradient descent is similar to a ball with no mass falling down a hill. Other methods (heavy-ball and Nesterov acceleration) are
62.
by skierscott 10y ago
> pro level products into thin clients I guess Apple is skating to where the puck is going to be. I have a MBP and any serious computation is done on either EC2 or the cluster downstairs (a fairly sizable cluster). I do think that they o
63.
by skierscott 10y ago
> But Apple has always missed the cue on one critical piece: access to customer data. And that is what Google has always been successful at. > When you buy an Android phone you sign in to your Google account. Behind the scenes, Google
64.
by skierscott 10y ago
I work on algorithms that can be applied to images, and was equally surprised when I saw a video called "Computer color is broken." I investigated and wrote a post called "Computer color is only kinda broken"[1]. This po
65.
by skierscott 10y ago
I'm a second year graduate student in CS and also debating getting an MBA. I currently study optimization and machine learning, and in the future I'd like to lead a team of people studying similar topics. Initially, my goal was to
66.
by skierscott 10y ago
There are many open ended questions here. There's more flexibility to sci Fi then I thought.
67.
by skierscott 10y ago
Whenever I've seen similar problems in real analysis, the wording of the question always leaves no solution as a possibility. "True or false? Can 1234 be rearranged such that...".
68.
by skierscott 10y ago
Regardless, when I learned this method in my linear algebra it inspired me to write a blog post [1] showing the derivation! This is a classic programming question, and it's refreshing to see a closed form solution. [1]: https:/&#x
69.
by skierscott 11y ago
This article describes what I'm doing right now. The academic group I am a part of have developed an adaptive or active machine learning system. This code is being used by the New Yorker to run their caption contest (example contest at
70.
by skierscott 11y ago
> Go2shell also makes iTerm2 slightly even better, just don't mistype it without that 's'. There's also OS X's "cdf" command to cd to the currently open Finder window (I think of "cdf" as shor
71.
by skierscott 11y ago
Spark has the same functionality as Mailbox, including "snooze till later".
72.
by skierscott 11y ago
Yup, same experience (also for data science). The biggest helper I've had so far is jupyter-vim-bindings[1] [1]: https://github.com/lambdalisue/jupyter-vim-binding
73.
by skierscott 11y ago
> And yet every time I try to get Matlab users to use something else (Python or Julia) the one thing they almost immediately complain about is the lack of a GUI IDE as good as the Matlab one. Spyder offers a similar environment. It has a
74.
by skierscott 11y ago
The Center for Open Science (COS) has developed tools to help this, the Open Science framework (OSF) [1]. COS hopes that the OSF will be useful and gently nudge users towards sharing data/preregistering/"open science" an
75.
Least squares and regularization(scottsievert.com)
13 points by skierscott 11y ago|0 comments
76.
Stepping from Matlab to Python(scottsievert.github.io)
2 points by skierscott 11y ago|0 comments
77.
by skierscott 11y ago
cvxpy[1] is a good example of this. The contraints are specific symbolically and cvxpy optimizes for you (assumes problem is a convex optimization problem etc) [1]: https://github.com/cvxgrp/cvxpy
78.
by skierscott 11y ago
If a company looks down on mental health issues, do you want to work there?
79.
by skierscott 11y ago
I am (and I believe most scientific users are too). I know that personally I have no incentive to switch; Python 2.7 has all the features I need. That will change with Python 3.5 and the matrix multiplication operator @. Is there any other
80.
by skierscott 11y ago
Agreed. Using the tools is a very different thing from knowing why they work. Additionally, I wonder how easily this can be adapted to other planes besides the 747. How much does it rely on cabin width/length?
81.
by skierscott 11y ago
The L0/L1 norm can have surprising benefits under certain assumptions. If you know (or assume) your signal is sparse (or mostly zero) in some domain, you're just trying to minimize the nonzero terms with the L0 norm.... But that
82.
by skierscott 11y ago
Nothing can exceed the speed of light. This equation represents that. I also seem to remember that causality is not preserved. This means that if A happens before B in one frame, it's possibly that B happens before A in another frame.
83.
by skierscott 11y ago
I'm a scientific user with no incentive to switch; I'm still on Python 2.7. But the release of the @ matrix multiplication operator in Python 3.5 gives me strong incentive to switch. I'll be switching as soon as Anaconda supp
84.
by skierscott 11y ago
Better how so? Faster? Math grads might be able to use some mathematical trick to dramatically speed up the code. Better modularity? Version control? Tests? Experience with SDKs?
85.
by skierscott 11y ago
Skiing. It's cool because it's * crazy addictive. If you can go you will * a decent workout. If I'm in a course or trying, even run is like a sprint * allows you to connect with others -- you can ride the chair up together! B
86.
Computer color is only kinda broken(scottsievert.github.io)
5 points by skierscott 11y ago|0 comments
87.
by skierscott 12y ago
Long story short, names are not consistent across socioeconomic groups. That likely plays a larger role than the actual name.
88.
by skierscott 12y ago
Understanding papers depends on what your field is. For example, I'm interested in optimization/signal processing/machine learning and finishing up my education. In the related journals I can barely understand the notation an
89.
N dimensions, linear algebra, linear functions and closed-form solutions(scottsievert.github.io)
7 points by skierscott 12y ago|0 comments
90.
by skierscott 12y ago
And more clicks.

More ›