I am making a project which is based on machine learning, and it uses a list of parameters as input. There are around 3000 such inputs. The algorithms learns the dataset every time I execute the code.
Is there any way, so that, I just run the program of learning the datasets once and use its parameters every time the code is run again?
That depends very much on your framework, but all the ones I know of do save those trained weights for you to use as a trained model. You need to look in the documentation for the framework you're using. The term for using a model to predict on new data is prediction, inference, or scoring.
For instance, in Caffe, you use the trained model something like this:
./build/tools/caffe test -model models/alexnet/train_val.prototxt
-weights alex_trained_iter_500000.caffemodel -iterations 1000
The critical part here is that caffemodel file used for weights.
Does that help?
I think what you're looking for is pickling, you can achieve that by doing something like this
from sklearn.externals import joblib
joblib.dump(clf, 'filename.pkl')
clf = joblib.load('filename.pkl')
You can either run your program once, save the pickle and then continue from there. Or check the existence of the pickle file each time the program runs and train the model if it doesn't exist.
from pathlib import Path
my_file = Path("/path/to/file")
if my_file.is_file():
References:
http://scikit-learn.org/stable/modules/model_persistence.html
