[Simra Temmy]

Thank you so much,
Best regards,
Simra

[Julia Crocco]

Hello,

I was able to install the brian command and open the plugin in terminal but when I try to run the code I get the error:

mach-o file, but is an incompatible architecture (have (x86_64), need (arm64e))

I have read about work around for Python but has anyone been able to overcome this on a Mac with an M1 chip?

Thanks!

Julia

[Daniel Schaefer]

Hi Julia,

Unfortunately, it sounds like the command uses some compiled code that is not compatible with your processor. One (less than ideal) solution was recently suggested by William Lisowski in another related thread. You could downgrade to the x86_64 version of Stata. This should force your mac to emulate the entire Stata application with Rosetta 2 - including the -brain- command. However, this will likely be computationally inefficient, which could be a serious hindrance depending on the amount of data you have and the number of parameters in your model. It might also cause some unexpected stability issues.

A second possible solution is to reach out to the original creator of the command to ask if they are willing to compile for the M1 chip. You could create a new issue on the Github. Looks like the code was last updated a couple of years ago, so it is unclear whether or not it is still being maintained. Assuming the source code on Github is the latest, it may also be possible to compile the code yourself with Clang. Theoretically, if the source is valid C, the compiler should be able to compile the code to assembly for your processor with a single terminal command. Practically speaking, I think it is likely that you will need to troubleshoot a few compiler errors before the code compiles. This could be quite difficult, particularly if you don't have much experience with C.

[Thorsten Doherr]

Hi Daniel,

thank you for your research in regard of the integration of the plugins into the Mac world. As I do not have access to such a device, there won't be any updates from my side regarding Mac OS (or Unix).
As you have already pointed out, the github page contains the source code for the plugins (brain.c supports multi-processing while brainsp.c is using a single processor). However, I was only able to compile the non-multiprocessor (sp) version on an older Mac because it seems that "openmp" is only properly supported by Windows.
Unfortunately, plugins are not only bound to the hardware/OS they are compiled for, you also need a STATA licence for all the different configurations to be able to test them.

[Minda Tan]

Hi everyone,

I have just recently uploaded a neural network module for Stata.

brain.ado is neural network application to provide some remedy for the distinct lack of such methods for Stata.

It is a no frills implementation of a backpropagation algorithm in Mata designed for a hassle free setup of multi-layered neural networks. After training the whole network can be saved/loaded using so called brain-files (default postfix .brn). The network is represented by a set of reserved matrices to provide transparent access to all components and to support older Stata versions. Additional functions facilitate the calculation of pseudo-marginal effects or signal through-put, but the main utility is of course prediction, i.e. for propensity scores or classification.

You can install it with:

ssc install brain

Cheers,
Thorsten

Thank you for your contribution! It is a great package for neural network.

I have a question about the coefficients of the inputs (the "x"s). Should I use "brain margin" or "brain signal" to obtain the coefficients? I don't completely understand the instruction on "brain signal".

[Thorsten Doherr]

You should use the margin command which works like its eponymous Stata command. It reports the difference in the 'y' (there can be multiple Ys) when the parameter signal of 'x' is set to zero (which is the minimum value for x before normalization). This is repeated over all paramters to report the "marginal effect" of each 'x'. This is not a coefficient as the functional form is a black box neither will you get standard errors.