ambrosys / glyph Goto Github PK

View Code? Open in Web Editor NEW

91.0 11.0 16.0 559 KB

a python 3 library based on deap providing abstraction layers for symbolic regression problems.

License: GNU Lesser General Public License v3.0

Makefile 0.17% Python 99.83%

symbolic-regression genetic-programming

glyph's People

Contributors

Stargazers

Watchers

Forkers

varunmittal91 vishalbelsare magnati sayin dfajardod ahmadiphy zeta1999 summuk omidr1370 wmoalxf sizzles riuza-c-137 lkampoli rishirelan jifengthu buaaguotong

glyph's Issues

Question: How to make use of optimized constants?

The package description mentions "interfacing constant optimization to scipy.optimize", but we have trouble finding any further documentation on this. How can we make sure our constants are optimized?

parallel factory

Clean additional Gooey attributes

Gooey is not going to use the callback functions, which were implemented.

They can be cleaned.

More improvements for experimental implementation

Smart termination of constant optimization. If it does not look like the fitness is improving fast enough (compared to the best fitness), give up.
Implement structural constants
Constraints based on simplified expressions (do not evaluate zero or inf functions)
Send an entire generation to the experiment to be evaluated at once (returns vector of fitness values).
Asynchronous communication in glyph_remote for sending/receiving functions and fitness values

Application.pareto_fronts is not being updated

glyph-remote: order of config parameters

argparse non defaults > file/server > argsparse-defaults

constant optimization

better interface for p0
refactor hillclimber

GUI logging

Logging options should be loaded before gui, e.g. glyph-remote --gui -vvv should enable debug logging to console to monitor gui startup.

please add tensorflow 2.0 keras

hi,

Do you like AI? im sure many users and researchers and developers would appreciate the ability to apply genetic programming to neural networks / computation graphs... but i find support for this seemingly obvious idea to be really lacking in 2019. This looks like a cool project. I would like to use it, but if we can't hook it up to our existing tensorflow / keras stuff, it seems we can't use Glyph!

How hard do you think it would be to add tf.keras support to glyph so we could make computation graphs that run fast with tensorflow and have strongly typed / SHAPE CHECKING (to avoid infinity bugs and frustration) for tensors?

glyph remote options

correct create_method nsga2 -> halfandhalf...

Experimental Implementation Improvements

Used configurable stop condition
Set maximum tree depth

glyph-remote: only schedule duplicates once

gui for parameter configuration

Add gui option --gui for glyph-remote.
Installation of gui option should be optional, e.g. pip install "pyglyph[gui]"

sympy-related functionality

Fail gracefully for simplification and NullSpace checking.
Issue warnings if primitive set in incompatible with sympy (e.g. Add/add).

Loggin optional in output file fuer glyph_remote

Initialisation of individuals of the first generation

Typo: Factoy

glyph/glyph/assessment.py

Line 14 in 343a075

class SingleProcessFactoy:

glyph-remote request

The cli should have a setter for the sent individual. Either one individual or a whole generation (or a certain number) shall be sent. this must be configurable.

Roadmap

v0.2

documentation

v0.3

refactor: constant optimization #9

v0.4

parallel factory #10
tests for glyph-remote

KeyboardInterrupt in glyph-remote

This should finish the current generation, shut down communication and make a checkpoint.

When running example lorenz.py Residuals are not finite in the initial point.

After doing a clean checkout of the glyph github project, I'm getting below error on running the Lorenz example.

Python 3.6.4

<deap.gp.PrimitiveSet object at 0x1069be320>
Traceback (most recent call last):
  File "examples/control/lorenz.py", line 137, in <module>
    main()
  File "examples/control/lorenz.py", line 78, in main
    app.run()
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/application.py", line 173, in run
    self.gp_runner.init(self.args.pop_size)
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/application.py", line 77, in init
    self._update()
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/application.py", line 87, in _update
    self._evals = self.assessment_runner(self.population)
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/assessment.py", line 53, in __call__
    for ind, fit in zip(invalid, fitnesses):
  File "examples/control/lorenz.py", line 47, in measure
    popt, rmse_opt = assessment.const_opt_leastsq(self.rmse, individual)
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/assessment.py", line 165, in const_opt_leastsq
    return const_opt(measure, individual, lsq=True, default_constants=default_constants, f_kwargs=f_kwargs, **kwargs)
  File "/Users/bartolkaruza/dev/proj/uva/glyph/glyph/assessment.py", line 152, in const_opt
    res = opt(fun=closure, x0=p0, **kwargs)
  File "/Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages/scipy/optimize/_lsq/least_squares.py", line 805, in least_squares
    raise ValueError("Residuals are not finite in the initial point.")
ValueError: Residuals are not finite in the initial point.

Make init output

Bartols-MBP:glyph bartolkaruza$ make init
pip install -r requirements.txt
Requirement already satisfied: cache.py==0.1.3 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 8))
Collecting deap==1.0.2.post2 (from -r requirements.txt (line 9))
  Using cached deap-1.0.2.post2.tar.gz
  Requested deap==1.0.2.post2 from https://pypi.python.org/packages/5b/d7/a49d3dd7aa8cbaf2b1ac8f4d6495824c886fea8b3dac4a73dc4df94cad76/deap-1.0.2.post2.tar.gz#md5=ccf5ed7562e4d6236c9416e3b5a9d941 (from -r requirements.txt (line 9)), but installing version 1.0.2
Requirement already satisfied: dill==0.2.5 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 10))
Requirement already satisfied: joblib==0.11 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 11))
Requirement already satisfied: mpmath==0.19 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 12))
Requirement already satisfied: numpy==1.11.2 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 13))
Requirement already satisfied: pyyaml==3.12 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 14))
Requirement already satisfied: pyzmq==16.0.2 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 15))
Requirement already satisfied: scipy==0.19.0 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 16))
Requirement already satisfied: sqlitedict==1.5.0 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 17))
Requirement already satisfied: sympy==1.0 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 18))
Requirement already satisfied: toolz==0.8.1 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from -r requirements.txt (line 19))
Installing collected packages: deap
  Found existing installation: deap 1.0.2
    Uninstalling deap-1.0.2:
      Successfully uninstalled deap-1.0.2
  Running setup.py install for deap ... done
Successfully installed deap-1.0.2
pip install -e .
Obtaining file:///Users/bartolkaruza/dev/proj/uva/glyph
Requirement already satisfied: deap in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: dill in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: numpy in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: pyyaml in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: scipy in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: sympy in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: toolz in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: pyzmq in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: cache.py==0.1.3 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: mpmath>=0.19 in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from sympy->pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: sqlitedict in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from cache.py==0.1.3->pyglyph==0.3.5+0.g2a2143f.dirty)
Requirement already satisfied: joblib in /Users/bartolkaruza/.pyenv/versions/3.6.4/lib/python3.6/site-packages (from cache.py==0.1.3->pyglyph==0.3.5+0.g2a2143f.dirty)
Installing collected packages: pyglyph
  Found existing installation: pyglyph 0.3.5+0.g2a2143f.dirty
    Uninstalling pyglyph-0.3.5+0.g2a2143f.dirty:
      Successfully uninstalled pyglyph-0.3.5+0.g2a2143f.dirty
  Running setup.py develop for pyglyph
Successfully installed pyglyph

tests for numerics module

readthedocs autodoc fails

readthedocs does not support python 3.5 yet, hence autodoc fails.

RuntimeError: generator raised StopIteration

Hi
I am learning to use this program, however, when running the lorenz.py example I get this error:

Traceback

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\control_problem.py", line 58, in odeint
raise StopIteration

StopIteration

The above exception was the direct cause of the following exception:

Traceback (most recent call last):

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\lorenz.py", line 137, in
main()

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\lorenz.py", line 77, in main
app.run()

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\application.py", line 184, in run
self.gp_runner.init(self.args.pop_size)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\application.py", line 79, in init
self._update()

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\application.py", line 89, in _update
self._evals = self.assessment_runner(self.population)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\assessment.py", line 56, in call
for ind, fit in zip(invalid, fitnesses):

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\lorenz.py", line 46, in measure
popt, rmse_opt = assessment.const_opt_leastsq(self.rmse, individual)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\deprecated\classic.py", line 241, in wrapper_function
return wrapped_(*args_, **kwargs_)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\assessment.py", line 177, in const_opt_leastsq
return const_opt(measure, individual, lsq=True, default_constants=default_constants, f_kwargs=f_kwargs, **kwargs)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\assessment.py", line 158, in const_opt
res = opt(fun=closure, x0=p0, **kwargs)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\scipy\optimize_lsq\least_squares.py", line 881, in least_squares
J0 = jac_wrapped(x0, f0)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\scipy\optimize_lsq\least_squares.py", line 875, in jac_wrapped
kwargs=kwargs, sparsity=jac_sparsity)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\scipy\optimize_numdiff.py", line 384, in approx_derivative
use_one_sided, method)

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\scipy\optimize_numdiff.py", line 454, in _dense_difference
df = fun(x) - f0

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\scipy\optimize_numdiff.py", line 348, in fun_wrapped
f = np.atleast_1d(fun(x, *args, **kwargs))

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\glyph\assessment.py", line 148, in closure
return measure(individual, *consts, **f_kwargs)

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\lorenz.py", line 54, in rmse
y = self.trajectory(individual, *f_args)

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\lorenz.py", line 62, in trajectory
return control_problem.integrate(dy, yinit=self.yinit, x=self.x, f_args=f_args)

File "C:\Users\81599\Documents\Python Scripts\Evolutionary Algorithms\Genetic Programming\Glyph\glyph-master\examples\control\control_problem.py", line 25, in integrate
y = np.vstack(res).T

File "<array_function internals>", line 6, in vstack

File "C:\ProgramData\Anaconda3\envs\glyph\lib\site-packages\numpy\core\shape_base.py", line 280, in vstack
arrs = atleast_2d(*tup)

RuntimeError: generator raised StopIteration

thanks in advance

raise warning if you cannot recover from a constraint

Instead of silently returning the input.
See constraints.py.

MultiOutput Regression Example

Hi,

I was wondering, does glyph support multioutput for symbolic regression? I can not seem to find an example for it.

Thanks

Examples are broken due to a class that was moved

Looks like AExpressionTree was moved from glyph.gp to glyph.gp.individual so some examples cannot start up and fail due to an import error.

Would it makes sense to at least import the examples files in the test suite to ensure they are up to date?

Failure of `make init`

Summary

On a fresh conda virtual environment, with glyph installed using pip install pyglyph, cloning the repos and running make init as per documentation instructions fail

Steps to reproduce

conda create -n glyph and conda activate glyph
conda install pip, pip install pyglyph
git clone https://github.com/Ambrosys/glyph
cd glyph
make init

Environment

What is the current bug behavior?

Installation script fails with error message:

Obtaining file:///Users/simenkva/Code/glyph_testing/glyph
  Installing build dependencies ... done
  Checking if build backend supports build_editable ... done
  Getting requirements to build editable ... error
  error: subprocess-exited-with-error

  × Getting requirements to build editable did not run successfully.
  │ exit code: 1
  ╰─> [34 lines of output]
      /Users/simenkva/Code/glyph_testing/glyph/versioneer.py:423: SyntaxWarning: invalid escape sequence '\s'
        LONG_VERSION_PY['git'] = '''
      Traceback (most recent call last):
        File "/Users/simenkva/anaconda3/envs/glyph/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 353, in <module>
          main()
        File "/Users/simenkva/anaconda3/envs/glyph/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 335, in main
          json_out['return_val'] = hook(**hook_input['kwargs'])
                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/Users/simenkva/anaconda3/envs/glyph/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 132, in get_requires_for_build_editable
          return hook(config_settings)
                 ^^^^^^^^^^^^^^^^^^^^^
        File "/private/var/folders/qc/b2y8zg713_x554g8ykkcbcbh0000gp/T/pip-build-env-dy00n4jd/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 448, in get_requires_for_build_editable
          return self.get_requires_for_build_wheel(config_settings)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/private/var/folders/qc/b2y8zg713_x554g8ykkcbcbh0000gp/T/pip-build-env-dy00n4jd/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 325, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=['wheel'])
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/private/var/folders/qc/b2y8zg713_x554g8ykkcbcbh0000gp/T/pip-build-env-dy00n4jd/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 295, in _get_build_requires
          self.run_setup()
        File "/private/var/folders/qc/b2y8zg713_x554g8ykkcbcbh0000gp/T/pip-build-env-dy00n4jd/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 487, in run_setup
          super().run_setup(setup_script=setup_script)
        File "/private/var/folders/qc/b2y8zg713_x554g8ykkcbcbh0000gp/T/pip-build-env-dy00n4jd/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 311, in run_setup
          exec(code, locals())
        File "<string>", line 29, in <module>
        File "/Users/simenkva/Code/glyph_testing/glyph/versioneer.py", line 1483, in get_version
          return get_versions()["version"]
                 ^^^^^^^^^^^^^^
        File "/Users/simenkva/Code/glyph_testing/glyph/versioneer.py", line 1415, in get_versions
          cfg = get_config_from_root(root)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/Users/simenkva/Code/glyph_testing/glyph/versioneer.py", line 344, in get_config_from_root
          parser = configparser.SafeConfigParser()
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      AttributeError: module 'configparser' has no attribute 'SafeConfigParser'. Did you mean: 'RawConfigParser'?
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

× Getting requirements to build editable did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.
make: *** [init] Error 1

Possible fixes

It seems the installation script uses a deprecated function from configparser.

Test resume in glyph remote

Use Github Actions for CI

GUI progress bar update

Progress bar of the gooey should be updated after each generation.

glyph.utils.numeric SlowConversionTerminator not working properly ?

Summary

Hi, as I'm about to use pyglyph for a science projet in school, I am currently going through your code and I think I found something wrong with the SlowConversionTerminator class. Sorry in advance if I'm not using github correctly, I'm new to it, but the concerned code si pretty short, so I hope to be clear enough.

If my understanding is correct, the whole desired effect of using this class is to optimize some function (find the values of its variables so that the result is close to 0), and killing the optimisation process if it detects that the optimisation is too slow, hence the name of the class.

As I first read through this code, I did not understand why the variable "fev" was constantly increasing, then I tested with the following code for myself :

The function used obviously cannot be optimized, I wanted to see if it detects it, but it did not. The process on the whole, just spams the current optimisation method (hill_climb here) until maxfev is reached : the maximum number of iteration (by default maxfev=1000 in this case).

The current if statement never could have been met (in the main loop of the class) :

As fev is constantly increasing, self.memory[fev] is always of length 1 (and self.min_stat is by default 10), and the "break early of the while loop" following this condition is never executed.

Possible fixes

I replaced the class code with the following, and it seems to work at least for the case I was testing for.

Constant optimization improvements

Do not consider unused constants, e.g. f(x, c0, c1) = c0*x
Improve caching, e.g. f(x, c0, c1) = c0*x is equivalent to f(x, c0, c1) = c1*x with swaped c0 and c1

Higher-level API

Hello,

I would like to thank you for this amazing project! However, I want to ask if you have thought about a higher-level API, such as classifiers and regressors similar to scikit-learn with fit and predict?

I found this idea in the gplearn project, which also does symbolic regression.

While your API looks very flexible, it also looks like it is quite verbose for a new and "light" user like me.

Would you be open to a pull request from me, which would let people just create a regressor with given settings (such as population, generations, allowed primitives), and call fit and predict on it, similar to gplearn?

ambrosys / glyph Goto Github PK

glyph's People

Contributors

Stargazers

Watchers

Forkers

glyph's Issues

Summary

Steps to reproduce

Environment

What is the current bug behavior?

Possible fixes

Summary

Possible fixes

Recommend Projects

Recommend Topics

Recommend Org