Groups 224 of 99+ julia-users › Modelling and Simulation using MATLAB Julia 8 posts by 4 authors Michael Fox 4/21/14 There's a MOOC starting tomorrow that promises to teach Modelling and Simulation using MATLAB. (Yes, they did misspell Modeling. The course is being offered by Iversity in partnership with Hochschule RheinMain - University of Applied Sciences in Germany.) I took a similar class 15 years ago and I thought it would be fun to brush up some skills and try doing the course in Julia instead. If anyone else wants in, sign up and let me know. We can have a study group where we road-test Julia. I have no idea what to expect but it sounds like the level is introductory which is fine with me since I expect to spend more time hunting Julia bugs than actual coursework. :) https://iversity.org/c/13?r=e3bf1 Jiahao Chen 4/21/14 "Modelling" is not a misspelling; it is merely the preferred orthography for most of the non-US Anglosphere. joelvan...@gmail.com 4/22/14 The description of the course suggests they will make substantial use of Simulink. Is there anything like that for Julia? Regardless of the blocks-and-wires UI, is there a Julia library that supports dataflow programming for systems with discrete and continuous dynamics? I'm curious to know the answer, since I have some s/w development ideas on this subject... On Monday, April 21, 2014 7:49:17 AM UTC-7, Michael Fox wrote: There's a MOOC starting tomorrow that promises to teach Modelling and Simulation using MATLAB. (Yes, they did misspell Modeling. The course is being offered by Iversity in partnership with Hochschule RheinMain - University of Applied Sciences in Germany.) I took a similar class 15 years ago and I thought it would be fun to brush up some skills and try doing the course in Julia instead. If anyone else wants in, sign up and let me know. We can have a study group where we road-test Julia. I have no idea what to expect but it sounds like the level is introductory which is fine with me since I expect to spend more time hunting Julia bugs than actual coursework. :) https://iversity.org/c/13?r=e3bf1 Michael Fox 4/22/14 Re: [julia-users] Re: Modelling and Simulation using MATLAB Julia This looks promising: https://github.com/BenLauwens/SimJulia.jl But I've learned to be wary of the package list. I would say part of the fun will be coming up with lightweight solutions to replace Matlab's heavyweight solutions. Some simple data-flow framework using coroutines can be developed on the spot if existing packages are found lacking. - show quoted text - -- - Michael Ivar Nesje 4/22/14 Re: [julia-users] Re: Modelling and Simulation using MATLAB Julia Especially you should be wary of packages that states that the last commit was authored 11 months ago. It is not necessarily a sign of bad quality, but you might find out that the author is not available to help debug errors, and you might not even get your Pull Requests reviewed and merged. The issue tracker is also a nice place to look for activity, and it has a 9 months old (apparent) typo that has not yet been fixed. There has been lots of deprecations since May last year, so I would recommend that you try using 0.2, because many deprecation warnings have been removed in the latest master. The good news is that the licence allows you to create a fork and then your fixed package can be registred in METADATA instead. - show quoted text - Joel VanderWerf 4/22/14 Re: [julia-users] Re: Modelling and Simulation using MATLAB Julia Ivar, Where do I find the 0.2 version you mention? The github master seems to be at 0.1.3: https://github.com/BenLauwens/SimJulia.jl/blob/master/VERSION. On 04/22/2014 12:24 PM, Ivar Nesje wrote: > Especially you should be wary of packages that states that the last > commit was authored 11 months ago. It is not necessarily a sign of bad > quality, but you might find out that the author is not available to help > debug errors, and you might not even get your Pull Requests reviewed and > merged. The issue tracker is also a nice place to look for activity, and > it has a 9 months old > (apparent) typo > that has not yet been fixed. There has been lots of deprecations since > May last year, so I would recommend that you try using 0.2, because many > deprecation warnings have been removed in the latest master. > > The good news is that the licence allows you to create a fork and then > your fixed package can be registred in METADATA instead. > > > kl. 20:49:15 UTC+2 tirsdag 22. april 2014 skrev Michael Fox følgende: > > This looks promising: https://github.com/BenLauwens/SimJulia.jl > > > But I've learned to be wary of the package list. I would say part of > the fun will be coming up with lightweight solutions to replace > Matlab's heavyweight solutions. Some simple data-flow framework > using coroutines can be developed on the spot if existing packages > are found lacking. > > > On Tue, Apr 22, 2014 at 9:59 AM, > wrote: > > > The description of the course suggests they will make > substantial use of Simulink. > > Is there anything like that for Julia? Regardless of the > blocks-and-wires UI, is there a Julia library that supports > dataflow programming for systems with discrete and > continuous dynamics? > > I'm curious to know the answer, since I have some s/w > development ideas on this subject... > > On Monday, April 21, 2014 7:49:17 AM UTC-7, Michael Fox wrote: > > There's a MOOC starting tomorrow that promises to teach > Modelling and Simulation using MATLAB. (Yes, they did > misspell Modeling. The course is being offered by Iversity > in partnership with Hochschule RheinMain - University of > Applied Sciences in Germany.) I took a similar class 15 > years ago and I thought it would be fun to brush up some > skills and try doing the course in Julia instead. If anyone > else wants in, sign up and let me know. We can have a study > group where we road-test Julia. I have no idea what to > expect but it sounds like the level is introductory which is > fine with me since I expect to spend more time hunting Julia > bugs than actual coursework. :) > > https://iversity.org/c/13?r=__e3bf1 - show quoted text - Ivar Nesje 4/22/14 Re: [julia-users] Re: Modelling and Simulation using MATLAB Julia I ment Julia version 0.2, I have not looked at SimJulia versions. Ivar - show quoted text - Joel VanderWerf 4/22/14 Re: [julia-users] Re: Modelling and Simulation using MATLAB Julia Michael, Thanks for the SimJulia link. Some good ideas in there, but also some things missing. (Algebraic equations? State transitions?) Be wary of "lightweight" solutions. They may have naive semantics, inherited from the operational semantics of the programming language they are written in. There are many issues in hybrid (discrete+continuous) simulation that need to be thought out carefully. For example, what is the order of execution during a parallel update of the discrete states in multiple components? Is this order a source of non-determinism in the outcome of the simulation? Suppose that the dataflow topology (i.e. the "wires") can change dynamically (it can't in Simulink, but you need this for mobile agent simulations). Then what is the semantics of those updates relative to guard checks and updates of other discrete state vars? How (and when) does a state transition impact the value of a variable that has been defined by an algebraic equation, if the new state imposes a new equation? These questions are really easy to skip over if the concurrency model is just inherited from the programming language's own model, whether thread, coroutine, task, fiber, process. A specialized concurrency model, using fine-grained phases within each discrete step, may be necessary to coordinate execution without giving up determinism. And there are issues around the continuous dataflow abstraction itself. How does internal integrator state (intermediate step values) interact with non-integration functions? The last time I looked at Simulink (maybe 2009), this state was not preserved across delay blocks, so integrator and delay blocks do not commute without some loss of accuracy. Returning to Julia and simulation... I'm interested in implementing a simulation engine for dynamic networks of hybrid automata. I've worked on this before in C[1] and Ruby+C[2], but Julia now seems to be exactly the right language for this kind of framework. Btw, I'll be at the SF Julia meetup tonight[3]. If anyone wants to talk, come find me! [1] http://gateway.path.berkeley.edu/SHIFT/publications.html (very old papers) [2] https://github.com/vjoel/redshift (ongoing work), http://confreaks.com/videos/165-rubyconf2009-dsls-code-generation-and-new-domains-for-ruby [3] http://www.meetup.com/Bay-Area-Julia-Users/events/164510912 On 04/22/2014 11:49 AM, Michael Fox wrote: > This looks promising: https://github.com/BenLauwens/SimJulia.jl > > But I've learned to be wary of the package list. I would say part of the > fun will be coming up with lightweight solutions to replace Matlab's > heavyweight solutions. Some simple data-flow framework using coroutines > can be developed on the spot if existing packages are found lacking. > > > On Tue, Apr 22, 2014 at 9:59 AM, > wrote: > > > The description of the course suggests they will make substantial > use of Simulink. > > Is there anything like that for Julia? Regardless of the > blocks-and-wires UI, is there a Julia library that supports dataflow > programming for systems with discrete and continuous dynamics? > > I'm curious to know the answer, since I have some s/w development > ideas on this subject... > > On Monday, April 21, 2014 7:49:17 AM UTC-7, Michael Fox wrote: > > There's a MOOC starting tomorrow that promises to teach > Modelling and Simulation using MATLAB. (Yes, they did misspell > Modeling. The course is being offered by Iversity in partnership > with Hochschule RheinMain - University of Applied Sciences in > Germany.) I took a similar class 15 years ago and I thought it > would be fun to brush up some skills and try doing the course in > Julia instead. If anyone else wants in, sign up and let me know. > We can have a study group where we road-test Julia. I have no > idea what to expect but it sounds like the level is introductory > which is fine with me since I expect to spend more time hunting > Julia bugs than actual coursework. :) > > https://iversity.org/c/13?r=__e3bf1 - show quoted text -