I failed to find a simple way to print out how many parameters in the model and proposals to print it in release notes did not find the warm welcome:
https://github.com/mozilla/DeepSpeech/issues/344
https://github.com/mozilla/DeepSpeech/issues/3392
How many parameters in DeepSpeech 0.8.2 and 0.9.0 models?
PS: thread on discourse.mozilla
Great question, but just for reference, as many people look at the answers first:
The English models have 47224861 parameters and the Chinese model in 0.9.0 has 47689984 parameters.
Expanding on that, DeepSpeech can best be measured by the n_hidden value which indicates the layer width of the model and the number of hours trained.
Usually, you get faster and better answers for DeepSpeech on their Discourse.
Related
I currently have several ARIMA models in R which I'm looking to present formally within a paper.
I'm just wondering if a tool exists that will allow me to illustrate these results, without having to manually extract this data?
Unless I'm mistaken, I understand stargazer does accommodate arima variations, but not the "Arima" code found within the forecast package. When I have tried to use the latter, I have received an error message telling me it is an "unrecognised object type".
Is there anything else that can be done?
Separately, for other types of diagnostic outputs (e.g. Augmented Dickey-Fuller tests, etc), are there other ways kable, etc, can be used to present these results?
Thanks in advance for any help!
It would be great if you could help me in understanding how randomization works in UVM and what factors affect it apart from ntb_random_seed . I am observing a testcase which is using same ntb_random_seed is following different randomization only with the addition of display macros. Is it possible? I hope my question is clear. Let me know, if you need additional information.
You need to read the 1800-2012 LRM section 18.14 on random stability. It lists the possible causes of instability. Older versions of the UVM had problems with the display macros, especially if you used things like the report catcher.
I'm looking for a tool to convert a SBML model into a Matlab function. I've tried SBMLTranslate() function from libSBML but this returns a Matlab struct, not a function. Does anybody know if such tool exists? Thanks
There are at least three efforts in this direction:
Frank Bergmann offers an online service for SBML translation where you can upload an SBML file and it will generate a MATLAB file. The comments at the top of the generated MATLAB file explain how to use the results. The C++ source code is available on SourceForge.
Bergmann's code referenced above was used by Stanley Gu to create sbml2matlab, a Windows standalone program. Off-hand, I don't know whether Gu's version changed or enhanced the algorithm used by the Bergmann version, but it seems likely. (Note: Gu now works at Google and does not maintain this code anymore, as far as I know.)
The Systems Biology Format Converter (SBFC) is a framework written principally by Nicolas Rodriguez; it includes a collection of converters, one of which is an SBML-to-MATLAB converter. This converter is written in Java.
I have not compared the results of the translators myself yet, so cannot speak to the differences or quality of output. If you try them and have any feedback to relate, please let the authors know. Knowing what has or hasn't worked for real users will help improve things in the future.
A final caveat is that all of these have been research projects, so make sure to set your expectations accordingly. (This is not a criticism of the authors; the authors are very good – I know most of them personally – but the reality of academic development work is that we all lack the time and resources to make these systems comprehensive, hardened, polished, and documented to the degree that we wish we could.)
I am trying to develop a simulation in OpenModelica of a flow that has a single substance that will be liquid or vapor. The Modelica.Media.Water models do have two phases, but are extremely complicated, and would be very hard to reproduce for a completely different substance.
I would like to find a simple example of a two phase medium that I can work from. There is a partial package TemplateMedium and a partial package PartialTwoPhaseMedium, but I don't see any examples of how to write a completely new Medium that can be in either of two phase.
If anyone can provide a simple example, or just a list of the minimum set of properties and equations that are required that would be extremely helpful as a starting point.
To address some of the question in comments:
I am just getting started on this model, so I am trying to understand the details of how the Media model is constructed, and what my specifics are included in the model versus what has to be added for each new media. I working with propylene, so there is good data available. This is one of the media that is included in CoolProp, so being able to use ExternalMedia and CoolProp would be very useful, but I believe that these are not yet working with OpenModelica, from a number of comments and bug reports.
Generally, your medium model can be written in Modelica or you can reuse an existing external library. Writing good medium models is a lot of work, so reusing existing libraries is usually a good idea. This is the approach taken by ExternalMedia (open source) or TILMedia (commercial).
If you are interested in an open-source workflow, ExternalMedia in combination with Coolprop is a reasonable decision. All three projects OpenModelica, ExternalMedia and CoolProp accept contributions from the community, so maybe you should help improving these instead of writing your own library. There is a lot of work going on already, I am unsure of the current status. Writing qualified bug reports (including steps to reproduce the problem) is also a very welcome way to contribute.
For some applications, it might be good to have the Medium model directly in Modelica. This is the approach taken by Modelica.Media (obviously), HelmholtzMedia and the commercial media libraries from XRG or Modelon (not 100% sure about that). There are some more implementations, but these are neither open source nor commercial, only information are e.g. conference papers.
The examples you can look at include the R134a medium from the MSL or the code from the HelmholtzMedia library.
Also, looking at the ExternalMedia implementation might help.
For fluids that cannot change phase, there are some good examples in the Annex60 library.
As you have a pure substance that can change phase, your new medium should extend from PartialTwoPhaseMedium.
PartialTwoPhaseMedium is partial, defining only what functions are there, but (mostly) not the algorithms of the functions.
You will have to write an algorithm for each and every function that is available in the interface and does not have an algorithm in order to be fully compatible.
For a start, you should implement at least one of the setState funtions, e.g. the setState_ph function.
Then later, implement at least one setSat function and the BaseProperties.
If you implement your own medium, you also have the choice of how to model it: Using the full multiparameter Helmholtz energy equation of state, a simpler equation of state like Peng-Robinson or other cubic EoS, some polynomials or splines, table-based methods like TTSE or SBTL and probably many more options.
I'm doing a research project on random forest algorithm. I have found numerous implementations of the algorithm but the main part of the code is often written in Fortran while I'm completely naive in it.
I have to edit the code, change the main parameters (like tree depth, num of feature variables, ...) and trace the algorithm's performance during each run.
Currently I'm using "Windows-Precompiled-RF_MexStandalone-v0.02-". The train and predict functions are matlab mex files and can not be opened or edited. Can anyone give me a piece of advice on what to do or is there a valid and completely matlab-based version of random forests.
I've read the randomforest-matlab carefully. The main training part unfortunately is a dll file. Through reading more, most of my wonders is now resolved. My question mainly was how to run several trees simultaneously.
Have you taken a look at these libraries?
Stochastic Bosque
randomforest-matlab
If you're doing a research project on it, the best thing is probably to implement the individual tree training yourself in C and then write Mex wrappers. I'd start with an ID3 tree (before attempting C4.5 for instance.) Then write the random forest code itself, which, once you write the tree code, isn't all that hard.
You'll:
learn a lot
be able to modify them as much as you like
eventually move on to exploring new areas with them
I've implemented them myself from scratch so I can help once you post some of your own code. But I don't think anybody on this site will write the code for you.
Will it take effort? Yes. Will you come out of it with more knowledge and ability than you had going in? Undoubtably.
There is a nice library in R called randomForest. It is based on the original implementation of Breiman in Fortran but it is now mainly recoded in C.
http://cran.r-project.org/web/packages/randomForest/index.html
The main parameters you talk about (tree depth, number of features to be tested, ...) are directly available.
Another library I would recommend is Weka. It is java based and lucid.Performance is slightly off though compared to R. The source code can be downloaded from http://www.cs.waikato.ac.nz/ml/weka/