This is a neural network implemented entirely in numpy to do sentiment analysis. It is a simple feed forward neural network. It also uses nltk to preprocess the data and eliminate stopwords. It is a small network, and can be run on my Macbook pro in less than 5 minutes.
The sentences for the task can be found in sentences.txt and the labels are in labels.txt.
This program also implements the perceptron algorithm.
To run the network with the best parameters and the saved data, do this. This will be pretty quick (it is on my Macbook Pro):
python2.7 neural_net.py --best_parameters='all_best_parameters_dictionary.gzip' --unique_words='all_unique_words.npy' --test_data='sentences.txt' --test_labels='labels.txt'
In order to train the network and run it, you can run this command, but it will take a bit longer (about 5 minutes for me) since you will train it.
python3 hw1.py --test_data='sentences.txt' --test_labels='labels.txt'