Comment Feed for brianbec

Re: Fall Fury, the PDF!

brianbec — Wed, 06 Feb 2013 15:01:05 GMT

can't I give this 6 stars?

posted by brianbec

Re: The Lambda Calculus, General Term Rewriting and Food Nutrition

brianbec — Tue, 10 Jul 2012 17:59:29 GMT

@Richard -- here you go feedback always welcome!

https://github.com/rebcabin/MathematicaScenarios

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Fri, 13 Jan 2012 02:56:00 GMT

2 days ago, SteffenZeidler wrote

Hi, Steffen -- I was after something a little deeper, such as a user-defined numerical class (sketching, now), say logarithmic doubles, in which operator * is mapped to arithmetic + on doubles. I couldn't find a way to make Viterbi generic over both user-supplied classes with operator overloads AND over built-in types, where operators are not the same kind of thing as operator overloads. I could wrapper all the built-in types with operator * that maps to * etc., but that's a lot of work. That's essentially implementing my own "Numeric Tower" over the CLR basic types, and it didn't seem worth it to me to get a corner case like logarithmic double. Maybe there *is* a way, I just couldn't find it. Now I'm going to look at Richard.Hein's link

Edit: just an aside about why logarithmic doubles might be valuable: multiplying a bunch of small probabilities over and over eventually underflows doubles. Bad. Sometimes better to add logarithms of probabilities to get the log of the product of the probabilities.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Mon, 02 Jan 2012 22:50:30 GMT

Just had another little thought -- Since Viterbi is modifying its internal variables V and Path as observations present themselves in OnNext, Viterbi is, itself a little monad. It could inherit from the state monad or just built up on its own, but the point is that the OnNext function on the surface would internally call "Bind" or "SelectMany" on the Viterbi monad, that is, OnNext would just call the LINQ provider of Viterbi! This might shrink and robustify the code even more.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Sat, 31 Dec 2011 18:42:11 GMT

@JoshRoss: or even a precomputed corpus from the same folks

http://blog.afterthedeadline.com/2010/07/20/after-the-deadline-bigram-corpus-our-gift-to-you/

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Fri, 30 Dec 2011 23:03:45 GMT

@Jan de Vaan: Yes, this looks like a promising direction. http://www.codeproject.com/KB/cs/genericnumerics.aspx

I also found out it's possible to go generic on numeric types in F# IFF the methods are INLINED. That's because the compiler just "pastes" the inlined code after resolving the generic type, and then the normal method lookup finds the appropriate operator implementations. Not sure if this approach is available in C#.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Fri, 30 Dec 2011 22:59:31 GMT

@JoshRoss: you will also need "transition probabilities," which you can get from digram frequencies (frequencies of word pairs). One really cool way to get these is to analyze free texts on project gutenberg (public-domain ebooks as plain text!).

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Fri, 30 Dec 2011 22:45:43 GMT

@PCB: yup, just for symmetry (euphemism for "copy-paste" code ContainsKey would have been better.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Thu, 29 Dec 2011 15:04:33 GMT

@fwaris:Very nice!

Anyone taking up my invitation to try a spelling corrector with this?

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Thu, 29 Dec 2011 15:01:42 GMT

@jordanterrell: Yes, yes, yes, PROFILE to find perf bottlenecks! I jumped straight from the art code to the opposite extreme of in-place mutation, but did not profile the art code, just assumed that spinning up closures in State was the culprit.

And 100% agree that the provability and reliability of monadic functional code is worth it in many (most?) circumstances. On the perf critical path, profile and compromise as needed, focusing on implementation innards.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Tue, 27 Dec 2011 01:27:09 GMT

@AdamSpeight2008:looks like Option is F#'s implementation of the Maybe monad -- different in name only, so far as I can see. They even have a good fraction of the LINQ operators implemented on it. I hope to be able to say more about F#, in general, after getting some more seat time with it.

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Tue, 27 Dec 2011 01:17:01 GMT

@Simon: that will work great if you have explicit data that you can represent in-line in the code (and it so happens I do in my Wikipedia example). But I don't see how to take advantage of these collection initializers when reading data from an external source. For an application like that, I can still Aggregate over a composable dictionary, as in lines 168 through 174 in my unit test DictionaryExtensionsTests.cs... like this

const int range = 1280;
var kvps = Enumerable
    .Range(0, range)
    .Select(i => new KeyValuePair(i, Convert.ToChar(i).ToString()))
    ;

dict = kvps.Aggregate(dict, (d, kvp) => d.AddUnconditionally(kvp));

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Mon, 26 Dec 2011 18:30:08 GMT

@exoteric:the blackboard style is a cheat try this (on Win7)

Windows key & "+" -- turn on built-in magnifier

Windows key & "-" -- shrink back down

Control-Alt "i" -- inverse video your whole world

posted by brianbec

Re: YOW! 2011: Joe Albahari - LINQ, LINQPad, and .NET Async (and a little Rx, too)

brianbec — Mon, 26 Dec 2011 05:35:03 GMT

@exoteric:Even without NuGet, it still pays off bigtime. I have some LinqPad scripts where i have integrated OpenCv (via PInvoke), QuickGraph (see codeplex), the SQL server Geomety and Geography data types, LINQ / Objects, LINQ / XML, LINQ / Rx, Task lib, all in a single script (not just to use them for sophomoric show-off fun, but because they were the most economical way to get the job done!). I took the effort to pull this all into Visual Studio (not really hard, mostly just rationalizing the namespaces), and by gum it all just works. <3 + + + +

posted by brianbec

Re: Brian Beckman: Hidden Markov Models, Viterbi Algorithm, LINQ, Rx and Higgs Boson

brianbec — Mon, 26 Dec 2011 04:46:41 GMT

@bryanedds : I've had a couple of good experiences with F# (simulations of card and dice games, just for fun) but I need some more "seat time" with it before weighing in. I'm particularly interested in particle filters (/slaps forehead), as well as Kalman filters (Unscented, Extended, and others) in reactive form. I can hammer them out with confidence in C# or take the deep dive and try them in F#. We'll see

My first job out of grad school involved "heavy industrial" Kalman in Fortran II at Jet Propulsion Lab, where we used them to track everything we could track in the solar system. Daily runs had around 1,000 parameters and 25 million observations, but that was decades ago, and that kind of thing can be easily done on a windows phone now (required liquid nitrogren back then Imagine the applications!

posted by brianbec

Re: YOW! 2011: Joe Albahari - LINQ, LINQPad, and .NET Async (and a little Rx, too)

brianbec — Sat, 24 Dec 2011 01:44:19 GMT

One of my colleagues said "With ordinary tools, you code and then test; with LinqPad, you test and then code." Use it in concert with the Visual-Studio Unit Test Framework (VSUTF) and you will be writing bulletproof code with unbelievable speed.

Technique: get your code working in LinqPad where the code-test-look cycle is fast and frictionless (LinqPad's Dump() + the charting tools in System.Windows.Forms.DataVisualization + plus the Sho-viz libraries are worth their weight in diamonds for code-test-look speed!). Copy all your test stuff into VSUTF and the target code into your VS projects and BAKE 'EM. Really great!

posted by brianbec

Re: E2E: Whiteboard Jam Session with Brian Beckman and Greg Meredith - Monads and Coordinate Systems

brianbec — Thu, 08 Jul 2010 14:34:05 GMT

You've got some interesting thoughts, there. You're talking about complex-valued functions on the complex plane. The domain of such functions is an unbroken plane, a simple manifold, but the co-domains form very interesting shapes: manifolds in their own right, and composing such functions brings up all kinds of interesting phenomena. Your color-mapping idea is great. The link below has some variations on this idea that you might find useful:

http://www.kfunigraz.ac.at/imawww/vqm/pages/complex/index.html

and the following is a good book on the analysis of complex functions:

http://usf.usfca.edu/vca//

posted by brianbec

Re: E2E: Brian Beckman and Erik Meijer - Co/Contravariance in Physics and Programming, 3 of n

brianbec — Wed, 26 May 2010 06:23:04 GMT

N2Cheval -- yes, you're right: spotting the "signature" of contravariance is totally sensitive to the directions of the arrows. We will have to do better to nail that down. Too squishy so far But free jazz is like that sometimes.

posted by brianbec

Re: E2E: Brian Beckman and Erik Meijer - Co/Contravariance in Physics and Programming, 3 of n

brianbec — Wed, 26 May 2010 06:20:57 GMT

Kyrae -- this is a good one and fits my hunch that monads and coordinate systems are similar, that is, the coordinate transforms from {X} to {Y} and back are similar to the monad transformations from IEnumerable to IObservable and back. I'll think about this some more.

posted by brianbec

Re: Brian Beckman: On Analog Computing, Beckman History and Life in the Universe Redux

brianbec — Fri, 26 Mar 2010 21:10:24 GMT

Yes, I'm still here, HeavensRevenge. BTW, loved the youtube on COCONUT -- very, very cool!

posted by brianbec

Re: Brian Beckman: On Analog Computing, Beckman History and Life in the Universe Redux

brianbec — Fri, 12 Mar 2010 23:02:44 GMT

F# would probably be the best way to go, now.

posted by brianbec

Re: Brian Beckman: On Analog Computing, Beckman History and Life in the Universe Redux

brianbec — Thu, 11 Mar 2010 19:15:36 GMT

I think there is a bi-directional correspondence between computation and energy transfers. If you think of computation as manipulations of symbols in the lambda calculus or in the pi calculus, then that involves clearing and storing "memory cells," usually represented as states of switches in a network. Ed Fredkin showed that you can't change states of memories without energy transfers (and the entropy growth that goes along with them, by the second law of thermodynamics!), so it's not possible to do computations without spending energy and growing the heat in the Universe!

posted by brianbec

Re: Brian Beckman: On Analog Computing, Beckman History and Life in the Universe Redux

brianbec — Thu, 11 Mar 2010 18:15:20 GMT

Hi Akopacsi -- The rough idea on time is this: Consider a path -- a 1-dimensional curve -- passing through points in space-time. Every point along that curve has a particular set of 4 coordinates: 3 space coordinates and 1 time coordinate, for any reasonable choice of coordinate systems. Now, parameterize that curve by the incremental distance along the curve: as you move from one point to another, you go a certain "distance" in 4-space, a distance measured by the "metric tensor," which is a generalization of the Pythagorean or Euclidean distance. Locally, that incremental distance is sqrt(dx^2 + dy^2 + dz^2 - dt^2) (notice the minus sign!). This distance measure is unique for a choice of metric tensor and is called the "proper time." It's a kind of cosmological average of proper times over the Hubble motion of galaxies along their curves that measures the age of the Universe backwards 13 or 16 billion years. Very rough idea, but hope that adds some clarity.

I'll take another look at "Rigs of Rods," one of my all-time favorite pieces of software!

posted by brianbec

Re: E2E: Brian Beckman and Erik Meijer - Co/Contravariance in Physics and Programming, 2 of 3

brianbec — Sat, 09 Jan 2010 04:44:57 GMT

Fantastic, Curt! thank you.

posted by brianbec

Re: E2E: Brian Beckman and Erik Meijer - Co/Contravariance in Physics and Programming, 2 of 3

brianbec — Wed, 06 Jan 2010 16:53:38 GMT

Erik and I have started cooking up the last batch on this topic. Will have something to video soon.

posted by brianbec