RMarcus

9 ms·

31.
Computer-generated lines with a human feel(rmarcus.info)
1 points by RMarcus 9y ago|0 comments
32.
by RMarcus 9y ago
I have no idea :). I don't think this approach is state of the art, but I think the advantages are that RFs are lightweight when compared to ANNs.
33.
by RMarcus 9y ago
I've done no experiment, but I predict a boosted forest would perform poorly here. Since the first tree learns the majority of the information (and each of the following trees only fits the errors, and are therefore less "meaningf
34.
by RMarcus 9y ago
Both good considerations! This approach can be used to build a kernel that measures the similarity of two items with respect to the labels learned by the forest. I don't doubt that the specific algorithm used to grow the forest. Luck
35.
by RMarcus 9y ago
These are all great questions. The simplicity of the kernels produced might make them very amenable to analysis. I'm not sure I have the technical skills to do it though...
36.
by RMarcus 9y ago
I agree the comparison could be a lot more fair. My goal was to show how adding in labels can give you better principle components, but I should've used an RBF kernel, or some kind of unsupervised kernel, instead of linear PCA. I reall
37.
by RMarcus 9y ago
I'm the author -- this is a technique to use a trained random forest as a kernel function for similar data. It can be used as a sort of primitive transfer learning, or just as a way to get a very high quality kernel for a (labeled) dat
38.
by RMarcus 9y ago
repl.it has had this capability for a while, although the embedded environment can be a little heavy. https://repl.it/site/docs/embed
39.
by RMarcus 10y ago
The defaults are acceptable, and will produce a symmetrically encrypted file that can be quickly decrypted on even low-powered ARM cores in a reasonable amount of time. These suggestions strike a different balance between protection and spe
40.
by RMarcus 10y ago
I've worked at Los Alamos for 6 years and I didn't know this existed. Pretty cool.
41.
by RMarcus 10y ago
>It’s the same reason that communism does not work nearly as well as capitalism — things work best when you give people freedom as individual actors. You said "capitalism" but you meant "liberalism," and you said &quo
42.
by RMarcus 10y ago
"AWS Batch dynamically provisions the optimal quantity and type of compute resources (e.g., CPU or memory optimized instances) based on the volume and specific resource requirements of the batch jobs submitted." Optimal is a p
43.
The Java (and Scala) type system is unsound(io.livecode.ch)
2 points by RMarcus 10y ago|0 comments
44.
by RMarcus 10y ago
I wrote a library to handle (many cases) of invalid JSON, motivated by a similar experience. https://github.com/RyanMarcus/dirty-json I'm on my phone now, but later today I'll test to see if it would have wor
45.
by RMarcus 10y ago
I know that the field of deep learning / machine learning generally moves too fast for researchers to target conferences or journals, but several of the citations in the PDF of this article are missing (LaTeX inserted [?]). (GRU, in ca
46.
A Revolution in Supercomputing Is Coming, from Los Alamos, NM(huffingtonpost.com)
1 points by RMarcus 10y ago|0 comments
47.
by RMarcus 10y ago
It's very interesting to see a lot of the traditional assumptions behind system design reserach change with hardware. In-memory databases, choosing O(n) cache-sensitive algorithms over O(log n) random access ones, and single nodes that
48.
by RMarcus 10y ago
I'm sure what the poster meant was the Pico wasn't free/libre software. edit: > because Pico's earlier license was not a free software license, since distribution of a modified version of the code was ambiguously forb
49.
by RMarcus 10y ago
Well, they have benchmarks! Against Vertica! Which is surprising, because I imagine Vertica's EULA forbids that. https://clickhouse.yandex/benchmark.html#["1000000000",["Cli... ]
50.
by RMarcus 10y ago
Vertica is a pretty large commercial column store with a lot of customers. Although I'm not sure if the Yandex folks got permission to publish benchmarks about it... (MonetDB is a classic research columnar database, and I believe MemSQ
51.
by RMarcus 11y ago
I'll admit I didn't spend a whole lot of time thinking of a name. I was originally going to go with "javascript-resolution-prover", but when I thought of "Vulcan" I thought it was too good not to use. To my gre
52.
by RMarcus 11y ago
Correct! Prolog uses a graph-based, backwards chaining algorithm to explore only the needed part of the search space to answer a query. My system explores an arbitrarily large portion of the search space (possibly all of it, in the case tha
53.
by RMarcus 11y ago
Hi everyone, I'm the author. I created this partially to help me understand the concepts in Norvig & Russell's AI book. I plan on using it in a larger project that requires logic-based reasoning after I add in a few search heu
54.
by RMarcus 11y ago
In Python3, you are really looking at... (func(a) for a in collection) map(func, collection) as equivalent. If you want a list (and not a generator), you would need to do this: [func(a) for a in collection] list(map(f
55.
by RMarcus 11y ago
I'd say that's more what cost functions are for, but that is neither here nor there. I think the biggest takeaways are... 1) Machine learning in large software projects is complex because decisions made by one algorithm can influe
56.
by RMarcus 11y ago
> The biggest problem with a procedural/explicit query is a dynamic system. Without a query planner, you don't have the luxury of a system rewriting your queries. When Table A ~ Table B, but then Table B >> Table A, your
57.
by RMarcus 11y ago
This is neat. Kind of reminds me of the built-in LRU cache decorator[1] (as the author points out). I'd be interested to know what kind of policies there were to control the cache size (will it just expand to fill up Redis / memca
58.
by RMarcus 11y ago
I think some examples would go a long way.
59.
by RMarcus 11y ago
Perhaps this has already been said and I missed it, but Facebook has the capability to be a pretty neat public key repository. You could have a pretty high degree of confidence that the public key associated with a Facebook account belonged
60.
by RMarcus 11y ago
Just a kind reminder to USfg employees and their contractors -- if you have CM on your personal computer, regardless of the source, you're liable as if you stole it.

More ›