siamese.ipynb

datget.py

import pickle
import itertools
import numpy as np
import tensorflow.keras as keras

with open('HackathonMetadata.pkl','rb') as ff: 
    metad = pickle.load(ff) 

with open('HackathonData_Homology0.pkl','rb') as ff: 
    homol = pickle.load(ff)

def pairs_generator(batch_size = 1024):
    iterator = itertools.combinations(range(homol['indata'].shape[0]), r=2)
    bl1, bl2 = [], []
    for e, i in enumerate(iterator):
        bl1.append(i[0])
        bl2.append(i[1])
        if e % batch_size == 0 and e != 0:
            yield [homol['indata'][bl1], homol['indata'][bl2]], homol['distdata'][bl1,bl2]
            bl1 = []
            bl2 = []

train_siamese.py

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras import backend as K
from tensorflow.keras.layers import Lambda

from datget import pairs_generator

input_shape = (328)

left_input = layers.Input(input_shape)
right_input = layers.Input(input_shape)
model = keras.Sequential()
model.add(layers.Dense(60, activation='sigmoid',))
model.add(layers.Dropout(0.2))
model.add(layers.Dense(2, activation='sigmoid'))
encoded_l = model(left_input)
encoded_r = model(right_input)
norm = keras.layers.LayerNormalization(axis=1)
encoded_l = norm(encoded_l)
encoded_r = norm(encoded_r)
dist_output = layers.Dot(axes=1)([encoded_l, encoded_r])
prediction = layers.Dense(1,activation='sigmoid')(dist_output)
siamese_net = keras.Model(inputs=[left_input,right_input], outputs=prediction)

optimizer = keras.optimizers.Adam(lr = 0.00006)
siamese_net.compile(loss="binary_crossentropy",optimizer=optimizer)

epochs = 100
siamese_net.fit_generator(pairs_generator(), steps_per_epoch=1000, epochs= epochs)