mlp_language_classifier
A multilayer perceptron which learns to classify languages from a wordlist. Includes a GUI to visualise neuron activity.
Requirements / plagiarism
- Uses scikit-learn's
MLPClassifier()class as a network - Tkinter for visualisation
- Wordlists kindly provided by http://www.gwicks.net/dictionaries.htm (English, French and German supplied)
Usage
- Either run the example scripts e.g. english_to_french_[load or train].py, or:
from network.mlp import Language_MLP
english_or_french = Language_MLP(language1='english',language2='french',load_wordlists=False,load_network=True,
save_network=False,balance_wordlists=True,hidden_layers=(20,20))
The above will train a network on words of up to 15 characters using a multi-layer perceptron with a single output neuron. It will train using wordlists, which it will expect in location "./wordlists/<language1>.txt" and "./wordlists/<language1>.txt"
