Implementation of NeuralFingerprint(https://arxiv.org/pdf/1509.09292.pdf)
- keras (with theano as backend)
- rdkit
from nfp_keras import NFPLayer
from nfp_keras.features import gen_feature_vector_from_sdf
from keras.models import Sequential
from keras.layers import Dense
data_x = gen_feature_vector_from_sdf('hogepiyo.sdf')
model = Sequential()
# you should pass "batch_input_shape".Since NFPLayer must know batch_size before model.compile.
model.add(NFPLayer(2048, 62, 5, batch_input_shape=(batch_size, 24586,)))
model.add(Dense(1))
....
- visualize
some codes in features.py are borrow from original implementation (https://github.com/HIPS/neural-fingerprint)