blt

60 ms·

721.
by blt 8y ago
I am glad I'm not the only one. I find the hidden state confusing, and I prefer my code to be in plain text files in a git repo.
722.
by blt 8y ago
For sure, this is one of the legitimate uses of runtime vtable-based polymorphism. But it's also unnecessary if you assume the user is able to recompile the program.
723.
by blt 8y ago
An underappreciated facet of matlab that makes it mistake-resistant: by default, all parameters are passed by value. Thus, all functions are pure / referentially transparent at their interface, even if mutation is used internally. This
724.
by blt 8y ago
If you have never seen one, I strongly recommend seeking out a real vector arcade machine. The monitor has a beautiful, otherworldly appearance that can't be captured in photos or emulation. Somehow the lack of the raster grid makes th
725.
by blt 8y ago
from what I can tell, Marc Levoy is the main reason why the Pixel has all these abilities.
726.
by blt 8y ago
If you have some ingredients you want to get rid of, make a bigger recipe and have leftovers. If you don't, don't.
727.
by blt 8y ago
But then people use excuses like "I have thick skin, I wouldn't be offended if someone else said the same thing to me, I prefer honesty over decorum", etc. Most often these people come from a privileged background where negat
728.
by blt 8y ago
Red blob is one of the best programming sites! I love the interactive demos and the clear writing style. They always strike a good balance between rigor and intuition.
729.
by blt 8y ago
I desperately want to write my ML code in a statically typed language. I hope one of these Swift-based projects takes off. But I am worried that they will not treat Linux as a first-class platform.
730.
by blt 8y ago
I agree quaternions are a good parameterization of SO(3). But they can still cause unexpected problems, e.g. any machine learning task involving quaternions will be made harder because the learned function must deal with the double coverin
731.
by blt 8y ago
There does not exist a R^3 parameterization of the 3d rotations without some problem. https://en.m.wikipedia.org/wiki/Charts_on_SO(3)#Parametrizat...
732.
by blt 8y ago
I think it's more like a meme. Countless other mathematical insights in the same time period have been at least as important and counterintuitive.
733.
by blt 8y ago
I really wish I knew what Ghostface says at the end of "Scary Hours"...
734.
by blt 8y ago
Apirori algorithm for finding frequently occurring subsets in a set of sets of objects. I find the algorithm beautiful because: 1) It takes advantage of a simple insight: all frequent subsets must be unions of other subsets that are at leas
735.
by blt 8y ago
what?? the solution set is infinite, how does it choose?
736.
by blt 8y ago
Meanwhile, forked repositories still aren't searchable.
737.
by blt 8y ago
but copying GC can be easily implemented with sbrk, or on the bare metal if no MMU is present. mpirstitz's GC uses mmap but could easily be replaced with sbrk. the main challenge of dynamic memory management is fragmentation, which is
738.
by blt 8y ago
Everything in your second sentence is true, but it seems like LLVM has been a huge boost. I think a lot more programmers are comfortable with parsing and ASTs than they are with the x86 instruction set. Not having to write IL->CPU codege
739.
by blt 8y ago
This is cool. I wrote a similar Lisp but got stuck debugging my copying GC. The interaction between the Lisp environment chain and the C stack is tricky. I prefer something like this over Norvig's Lisp interpreter written in Python. No
740.
Reverse engineering a simple CMOS chip(youtube.com)
2 points by blt 8y ago|0 comments
741.
by blt 8y ago
Opening up with the claim that computer software is "based on the ideas of Claude Shannon" is strange. Does anyone seriously believe that Shannon's contributions to the foundations of computing stand so far above others'
742.
by blt 8y ago
Non sequitur, but in my experience the parent pointer in a tree can almost always be passed as an argument in whatever recursive function is traversing the tree.
743.
by blt 8y ago
It's wrong even if their model doesn't output a certainty (not all classifiers do). Almost all ML algorithms optimize the expected classification error under the training distribution . So if the training data contains 90% men, i
744.
by blt 8y ago
Non sequitur, but is Cody Cutler (you?) related to Dave Cutler, the designer of VMS and Windows NT? Maybe OS programming runs in the family :)
745.
by blt 8y ago
> An unoptimized research kernel The authors would probably not call it "unoptimized". From the last paragraph in Section 8.8: "It took effort to make Biscuit achieve this level of performance. Most of the work was in un
746.
by blt 8y ago
Overleaf makes you pay for the history feature, and for more than one collaborator. Not good for long term collaborative papers.
747.
by blt 8y ago
Wow, I did not know IDEs were making suggestions like this.
748.
by blt 8y ago
Good point. Jira's depth and customizability must be like candy for terrible micromanagers. Kind of like how C++ gets an exaggerated bad rap because it is really attractive to people who like building complicated, baroque systems.
749.
by blt 8y ago
> Supply and demand is volatile, so just because a unit averages $x does not mean it can't sometimes fetch 5x (see Uber's surge pricing). Smart pricing never does this, so I suspect there is no notion of demand surges built in
750.
by blt 8y ago
Basically, yes. Deep RL has the potential to make high accuracy/repeatability unnecessary, but it's still in the research lab setting.

More ›