在 matlab 中直接在 python 中使用 sklearn

Question

Please do not mark this as a duplicate of how to call python and sklearn from matlab? as the question is in my opinion not really answered.
Since Matlab Release R2014b I think, it is possible to directly use python from matlab .
In short words, you only have to place py in front of the python call.
I my setup (after giving matlab the python path with the command pyversion('PATH_TO_PYTHON') , that is running fine. I can even use dask multiprocessing. Quite cool. For example, executing py.dask.distributed.Client results in

  Python Client with properties:

              asynchronous: 0
                   cluster: [1×1 py.distributed.deploy.local.LocalCluster]
         get_futures_error: [1×1 py.method]
                coroutines: [1×1 py.list]
            scheduler_file: [1×1 py.NoneType]
                      loop: [1×1 py.tornado.platform.select.SelectIOLoop]
    recreate_error_locally: [1×1 py.method]
                  refcount: [1×1 py.collections.defaultdict]
                extensions: [1×1 py.dict]
                 scheduler: [1×1 py.distributed.core.rpc]
                       rpc: [1×1 py.distributed.core.ConnectionPool]
                   futures: [1×1 py.dict]
            scheduler_comm: [1×1 py.distributed.batched.BatchedSend]
                    status: [1×7 py.str]
           connection_args: [1×1 py.dict]
                        id: [1×43 py.str]
                generation: [1×1 py.int]
                   io_loop: [1×1 py.tornado.platform.select.SelectIOLoop]
                  security: [1×1 py.distributed.security.Security]

    <Client: scheduler='tcp://127.0.0.1:59795' processes=4 cores=4>

Coming back to the question: I have sklearn installed and can use it from the referenced python Installation. It is working the same way as dask. But MATLAB R2017a is not able to find sklearn.
A similiar call to the given above py.sklearn.cluster.dbscan results in

Undefined variable "py" or class "py.sklearn.cluster.dbscan".

Is there any python expert being able to explain?

Answer 1

I got a solution from the mathworks Support .
It reads the way, that maybe the python environment is not completely setup. I was asked to start matlab from within the Anaconda Prompt which has that complete arranged environment. Running matlab from there yielded the wanted results thus being able to use for example sklearn.
Further comparing the diffenrences from there showed up, that some more directories from python have to be added to the systems search path.

Further I learned, that running py.importlib.import_module(<MODULENAME>) will show details if that python module and its dependencies are available or not.

Answer 2

On a Mac:

• Open a new terminal window;

• type: which python (to find out where the default version of python is installed);

• Restart MATLAB;

• type: pyversion('/anaconda2/bin/python') , in the command line (obviously replace with your path).
• You can now run all the libraries in your default python installation.

For example:

>>py.sys.version;

>>py.sklearn.cluster.dbscan

Answer 3

You can create a conda environment and use it from MATLAB as follows. Note that it you are debugging Python at the same time, and making changes to some_awesome_python_module , you have to reload it every time (code below starting with clear classes is how to do that):

py_root_useFromMATLAB = fileparts(C:\anaconda_3\envs\useFromMATLAB\_conda.exe);
ENV = getenv('PATH');
ENV = strsplit(ENV, ';');
items_to_add_to_path = {
    fullfile(py_root_useFromMATLAB, 'Library', 'mingw-w64', 'bin')
    fullfile(py_root_useFromMATLAB, 'Library', 'usr', 'bin')
    fullfile(py_root_useFromMATLAB, 'Library', 'bin')
    fullfile(py_root_useFromMATLAB, 'Scripts')
    };
ENV = [items_to_add_to_path(:); ENV(:)];
ENV = unique(ENV, 'stable');
ENV = strjoin(ENV, ';');
setenv('PATH', ENV);
clear classes
module_to_load = 'some_awesome_python_module';
python_module_to_use = py.importlib.import_module(module_to_load);
py.importlib.reload(python_module_to_use);

Now you can use it like:

output = py.some_awesome_python_module.some_awesome_python_function(input)