Menu
▾
▴
Re: [Obi-plan-branch] [Obi-datatrfm-branch] plan and data_transformation
|
From: Philippe <ro...@eb...> - 2008-01-23 10:48:24
|
Hi Franck, All The document is indeed extremely useful (even if I have to admit it feels a bit like a show-stopper).Thanks for pointing it out (again). The definition for algorithm (if taken from your document) would therefore be: an algorithm is a plan which specifies how to compute the output (data) of a (mathematical) function knowing the input (data). Following this definition (and if agreed by all) would force us to revise our current definitions in order to identify the mathematical functions when possible. I say 'when possible' as it seems that some of 'algorithm-related' terms we have gathered identify 'Families' of algorithms based on their possible applications. In other cases, it seems to be more straightforward : So for example, if we take Cooley and Tukey Fast Fourier Algorithm,the function we can identify is a Fourier Transform (a linear transformation). A few basic mathematical functions were initially included in the Data_transformation branch (which specifies input and output) but were deemed to be outside the scope and mandate of OBI. I feel that performing the mapping you envisage between algorithm,function, type looks like an entire project on its own and pushes the envelop of OBI. Right now, I believe that we are looking for 'labels' to tag methods applied during data processing, more a CV than an ontology. It would be actually nice to have more input from your side on this topic (I mean a worked example that could be imported in OBI for instance) Cheers Philippe frank gibson wrote: >Hi > >This document, authored by Matt, may help on the discussion of algorithm >http://docs.google.com/Doc?id=dhkzc9hp_0tx33wc > >Frank > >On Jan 16, 2008 10:18 AM, Philippe <ro...@eb...> wrote: > > >>Hi Melanie, data-transformers >> >> >> >>>We are trying to see what is the progress regarding our action items >>>from the data transformation workshop in november. >>>Regarding the plan branch there was the addition of the >>>algorithm-related terms and the relation "utilizes" >>> >>>You acknowledged reception of the algorithm terms, so can I assume >>>there was no issue with them? >>> >>> >>I indeed acknowledged reception of the terms but your assumption does >>not hold true ;). >>Algorithm is a bit thorny and the definition reflect a variety of usage >>which prevent direct inclusion without incuring the risk of multiple >>inheritance. >>Also many definitions provided by the data analysis branch rely on >>absent parent terms such 'meta-algorithm' or meta-learning algorithm >>which are not defined. >> >>What is missing is a definition of algorithm but from the cases you have >>submitted I guess we can outline 2 properties: algorithm have a purpose >>(search, classification) and may have a method or a princinple (eg >>Cooley of Tuckey method for Fast Fourier Transformation). But just like >>other kind of plans, there are granularity issues,since for the case of >>Cooley and Tukey Fast Fourier Algorithm >>it could be considered as plan to carry out another plan. >> >>Another issue is that the same algorithm maybe used in a variety of >>applications. So how do we allow for this using the tools that we >>currently have ? This is unclear to me. >>Placing directly the terms as they are as children of algorithm isn't >>really clean. >> >> >> >>>I'm not trying to say "why are they not yet in the owl file", as we >>>also have in data transformation terms that still need to be added >>>even if they have been submitted a while ago, I just want to make sure >>>that on our side it's ok, and we completely trust you to add the terms >>>later on. >>> >>> >>That's jsut a fair question to ask but I hope you understand why they >>are not there yet. It also brings up the question of how to use >>relations and how to define/constraint those in keeping with other >>entities of OBI. >>An earlier post asking how to form relations and use them flared up but >>I did not have time to make a summary of the discussion and digest the >>outcome. >>But this is an issue that has been taunting us since the very beginning >>of OBI work and I hope that we will be able to outline some kind of >>methodology,understood and agreed upon by all obi developers during our >>meeting in Vancouver. These examples could be used for demonstration >>purposes. >> >>Thanks again for the head-up >> >> >> >>>Regarding the utilizes relation there has been a bit of discussion on >>>the dev list, but I don't think it actually resulted in any agreement? >>>I just sent an email on the obi-owl list to have more info. >>> >>> >>> >>I have attaching the terms submitted so far by the data transformation >>branch. >> >>AI: Pass to the plan branch: Genetic >>Algorithm, Boosting algorithm, Bootstrap algorithm, Weighted Majority >>Algorithm, Cooley and Tukey Fast Fourier Algorithm >> >>Evolutionary Algorithm (term added to be a parent of Genetic Algorithm) >>An evolutionary Algorithm is a generic population-based metaheuristic >>optimization algorithm. The methodology is described at >>http://en.wikipedia.org/wiki/Evolutionary_algorithm >> >>Genetic Algorithm >>A genetic algorithm (GA) is a search technique used in computing to find >>true or approximate solutions to optimization and search problems source. >>The methodology is described at >>http://en.wikipedia.org/wiki/Genetic_algorithm >> >>Boosting Algorithm >>A boosting Algorithm is a machine learning meta-algorithm for performing >>supervised learning. Boosting occurs in stages, by incrementally adding to >>the current learned function. At every stage, a weak learner (i.e., one >>that has an accuracy greater than chance) is trained with the data. The >>output of the weak learner is then added to the learned function, with >>some strength (proportional to how accurate the weak learner is). Then, >>the data is reweighted: examples that the current learned function gets >>wrong are "boosted" in importance, so that future weak learners will >>attempt to fix the errors. The methodology is described at >>http://en.wikipedia.org/wiki/Boosting >> >>Bootsrap Aggregating (name changed from Bootstrap Algorithm to be >>consistent with he definition and the commonly used term) >>bootstrap aggregating is a meta-algorithm to improve classification and >>regression models in terms of stability and classification accuracy. >>The methodology is described at >>http://en.wikipedia.org/wiki/Bootstrap_aggregating >>add bagging as alternative term. >> >> >>Weighted Majority Algorithm >>In machine learning, Weighted Majority Algorithm (WMA) is a meta-learning >>algorithm used to construct a compound algorithm from a pool of prediction >>algorithms, which could be any type of learning algorithms, classifiers, >>or even real human experts. The algorithm assumes that we have no prior >>knowledge about the accuracy of the algorithms in the pool, but there are >>sufficient reasons to believe that one or more will perform well. >>The methodology is described at >>http://en.wikipedia.org/wiki/Weighted_Majority_Algorithm >> >>Fast Fourier Transform (added to be a parent of Cooley >>and Tukey) >>Fast Fourier Transform is an efficient algorithm to compute the discrete >>Fourier transform (DFT) and its inverse. >>add alternative term FFT >>The methodology is described at >>http://en.wikipedia.org/wiki/Fast_Fourier_transform >> >>Cooley and Tukey Fast Fourier Algorithm >>The Cooley-Tukey algorithm is the most common fast Fourier transform (FFT) >>algorithm. It re-expresses the discrete Fourier transform (DFT) of an >>arbitrary composite size N = N1N2 in terms of smaller DFTs of sizes N1 and >>N2, recursively, in order to reduce the computation time to O(N log N) for >>highly-composite N (smooth numbers). >>It works significantly faster than other Fourier algorithm provided the >>size of the problem is a power of 2. >>The methodology is described at >>http://en.wikipedia.org/wiki/Cooley-Tukey_FFT_algorithm >> >> >>------------------------------------------------------------------------- >>This SF.net email is sponsored by: Microsoft >>Defy all challenges. Microsoft(R) Visual Studio 2008. >>http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ >>_______________________________________________ >>Obi-datatrfm-branch mailing list >>Obi...@li... >>https://lists.sourceforge.net/lists/listinfo/obi-datatrfm-branch >> >> >> > > > > > |
