wesleybeckner / salty Goto Github PK

View Code? Open in Web Editor NEW

10.0 3.0 2.0 543.03 MB

An interactive data exploration tool for json structured ionic liquid data from ILThermo (NIST)

License: MIT License

Python 2.25% Makefile 0.03% Ruby 0.01% Jupyter Notebook 97.71%

machine-learning sklearn molecular-design

salty's Introduction

SALTY

Salty is an interactive data exploration tool for ionic liquid data from ILThermo (NIST)

Installation

Dependencies

Salty requires:

python (>= 3.6)
scikit-learn (>= 0.19.1)
rdkit (>= 2017.09.1)

To take full advantage of rdkit you will also need Matplotlib >= 1.3.1.

User installation

You will first need to install rdkit:

conda create -n py36 python=3.6 anaconda
# activate the new virtual environment, e.g. on OSX/Linux
source activate py36
# on Windows
# activate py36
conda install -c rdkit rdkit

Salty can then be installed with:

pip install salty-ilthermo

Alternative versions of rdkit:

Should rdkit fail to install properly, refer to the Anaconda rdkit package entry and try:

#Recommended
conda install -c rdkit/label/beta rdkit

conda install -c rdkit/label/attic rdkit

conda install -c rdkit/label/testing rdkit

Development

Salty is currently underdevelopment by researchers at the University of Washington. Our research page can be found here.

Testing

After installation, you can launch the test suite from outside the source directory (you will need to have the pytest package installed):

pytest salty

salty's People

Contributors

Stargazers

Watchers

Forkers

jameslee97 stephco83

salty's Issues

Cannot import Imputer

When importing the package the following error is produced:

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
c:\Users\Holy Grail\Desktop\Science\il_properties.py in <cell line: 2>()
      [1](file:///c%3A/Users/Holy%20Grail/Desktop/Science/il_properties.py?line=0) # %%
----> [2](file:///c%3A/Users/Holy%20Grail/Desktop/Science/il_properties.py?line=1) import salty

File c:\Users\Holy Grail\.conda\envs\p4env\lib\site-packages\salty\__init__.py:3, in <module>
      [1](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/__init__.py?line=0) from __future__ import absolute_import, division, print_function
      [2](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/__init__.py?line=1) from .version import __version__ # noqa
----> [3](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/__init__.py?line=2) from .core import * # noqa
      [4](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/__init__.py?line=3) from .visualization import *

File c:\Users\Holy Grail\.conda\envs\p4env\lib\site-packages\salty\core.py:11, in <module>
      [9](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/core.py?line=8) from math import log
     [10](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/core.py?line=9) from math import exp
---> [11](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/core.py?line=10) from sklearn.preprocessing import Imputer
     [12](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/core.py?line=11) import numpy as np
     [13](file:///c%3A/Users/Holy%20Grail/.conda/envs/p4env/lib/site-packages/salty/core.py?line=12) from sklearn.preprocessing import StandardScaler

ImportError: cannot import name 'Imputer' from 'sklearn.preprocessing' (c:\Users\Holy Grail\.conda\envs\p4env\lib\site-packages\sklearn\preprocessing\__init__.py)

I believe import Imputer has been removed from later versions of scikit-learn, could you update either the code or the requirements.txt?

a new method will be part of a new submodule (adaptive_learn) dedicated to assigning GA targets and updating the qspr model. The method should identify hypothetical (target) property combinations outside the convex hull of training data.

visual api

dropdown menus for axes and groupby color.
assets directory for MPL images of cation/anion.
tooltip updates subplot of MPL image
timeline for adaptive_learn iterations?
something for most common substructure?

aggregate repeat data

There are many duplicate experiments in the ILthermo data files (same salt property at same T, P, done by different experimentalists). Would like to provide a function for merging these property values and recording the standard deviation. An additional function might consider this std during model fitting

agreggate_data() misslabels T and P

The df operations are correct but the output labels for Data_summary are swapped between T and P. should be an easy fix

create standard visualization tool for trained models

tool should automatically detect the number of outputs in a trained model and return a standardized figure (e.g. small multiple of parity plots, embedded score or RAAD, etc). Create as part of visualization.py module.