Reactive Programming in Machine Learning

bob zhang

22 Jul 2011 22 Jul '11

2:30 p.m.

Hi all, I am doing a survey on combining Functional Reactive Programming and Machine Learning. Has anyone did relevant research on this topic? Any discussion or link is appreciable. Best,bob

Show replies by date

Edward Amsden

22 Jul 22 Jul

3:28 p.m.

I did a survey of functional reactive programming, though there's no reference to machine learning: http://blog.edwardamsden.com/2011/05/survey-of-functional-reactive.html On Fri, Jul 22, 2011 at 2:30 PM, bob zhang wrote:

...

Hi all, I am doing a survey on combining Functional Reactive Programming and Machine Learning. Has anyone did relevant research on this topic? Any discussion or link is appreciable. Best,bob

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

-- Edward Amsden Student Computer Science Rochester Institute of Technology www.edwardamsden.com

bob zhang

4:01 p.m.

Thank your for kind help :-) 于 11-7-22 下午3:28, Edward Amsden 写道:

...

I did a survey of functional reactive programming, though there's no reference to machine learning:

http://blog.edwardamsden.com/2011/05/survey-of-functional-reactive.html

On Fri, Jul 22, 2011 at 2:30 PM, bob zhang wrote:

...
Hi all, I am doing a survey on combining Functional Reactive Programming and Machine Learning. Has anyone did relevant research on this topic? Any discussion or link is appreciable. Best,bob

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Tom Nielsen

23 Jul 23 Jul

2:06 p.m.

As far as I am aware, there has been very little work on combining these two, but that does not mean that it is a bad idea. I can give you some pointers from a very personal perspective: -Machine learning is mostly kernel methods and probabilistic inference, I can't really say much about how one would combine kernel methods and FRP, but more about probabilistic inference. -I work in physiology and we use FRP for data acquisition and for classifying different kinds of evidence. It appears that most data -- observed or inferred -- can be described as signals or as events. We have a paper about this but it is not yet published. -We do data analysis by building probabilistic models (i.e. probability distributions) for the "reactive" objects, i.e. signals or events, and applying bayesian inference to learn the parameters of these models. Lots of this work consists of thinking about what would be a good probability distribution for a signal or an event. -Gaussian processes make very good models for real-valued signals, especially when your underlying model is a stochastic differential equation (some of which can be rewritten as gaussian processes). -for events we use point processes such as the poisson processes. -I would love to work on probabilistic reactive control, but don't really have the time. You could use sequential monte carlo/particle filters to iteratively estimate the value of unobserved time-varying quantities and the use FRP-like systems to wire this up to an output signal. If you want to learn more about sequential monte carlo, there are lots of videos on videolectures.net. Nando de Freitas has a good introduction. Tom On Sat, Jul 23, 2011 at 2:30 AM, bob zhang wrote:

...

Hi all, I am doing a survey on combining Functional Reactive Programming and Machine Learning. Has anyone did relevant research on this topic? Any discussion or link is appreciable. Best,bob

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

David Barbour

2:50 p.m.

I have spent several months studying use of generative grammars in multi-agent reactive systems [1] - granted, not FRP in particular, but RDP is reasonably close [2]. This result is, implicitly, a distributed, federated machine-learning system (briefly described at [3]). The 'learning' supports rapid agreement between agents and eliminating the volatility seen in a stateless reactive model. Individual agents are simple; intelligence in this model emerges only as we scale. It's a very simple learning model: individually, agents try to generate grammars that that are (based on history) likely to be accepted by other agents. Developers control what can be learned by specifying a non-deterministic choice in the generative grammar (i.e. non-determinism is seen as 'permission to choose and learn', not 'random'). By a simple extension of grammars with time (i.e. a grammar generates a sentence that says not just what to do, but when to do it), I believe I can achieve a huge level of intelligent coordination and cooperation between agents. I.e. they'll automatically schedule their activities, and reactively adjust to accommodate changes in plans or the introduction of new agents. I tabled further study of this promising model until I sufficiently develop RDP, which is far more suitable than FRP for open, scalable systems. [1] http://lambda-the-ultimate.org/node/4012 [2] http://awelonblue.wordpress.com/2011/05/21/comparing-frp-to-rdp/ [3] http://lambda-the-ultimate.org/node/4012#comment-62877 http://lambda-the-ultimate.org/node/4012#comment-63105 On Fri, Jul 22, 2011 at 11:30 AM, bob zhang wrote:

...

Hi all, I am doing a survey on combining Functional Reactive Programming and Machine Learning. Has anyone did relevant research on this topic? Any discussion or link is appreciable. Best,bob

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

5052

Age (days ago)

5053

Last active (days ago)

List overview

Download

4 comments

4 participants

participants (4)

bob zhang
David Barbour
Edward Amsden
Tom Nielsen