sbalnojan

from torchbiggraph.config import parse_config
import attr
train_config = parse_config(CONFIG_PATH)

train_path = [convert_path(os.path.join(DATA_DIR, FILENAMES['train']))]
train_config = attr.evolve(train_config, edge_paths=train_path)

from torchbiggraph.train import train
train(train_config)

sbalnojan

edge_paths = [os.path.join(DATA_DIR, name) for name in FILENAMES.values()]

from torchbiggraph.converters.import_from_tsv import convert_input_data

convert_input_data(
    CONFIG_PATH,
    edge_paths,
    lhs_col=0,
    rhs_col=1,
    rel_col=None,

sbalnojan

import os
import random

def convert_path(fname):
    basename, _ = os.path.splitext(fname)
    out_dir = basename + '_partitioned'
    return out_dir

def random_split_file(fpath):
    root = os.path.dirname(fpath)

sbalnojan

graph_conv_filters = np.concatenate([A_norm, np.matmul(A_norm, A_norm)], axis=0)
graph_conv_filters = K.constant(graph_conv_filters)
num_filters = 2

model = Sequential()
model.add(GraphCNN(Y.shape[1], num_filters, graph_conv_filters, input_shape=(X.shape[1],), activation='elu',
                       kernel_regularizer=l2(5e-4)))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.01), metrics=['acc'])
model.summary()

sbalnojan

print("this filter includes the edges, so it should perform considerably better than before.:")
graph_conv_filters = A_norm
graph_conv_filters = K.constant(graph_conv_filters)
num_filters = 1

model = Sequential()
model.add(GraphCNN(Y.shape[1], num_filters, graph_conv_filters, input_shape=(X.shape[1],), activation='elu',
                       kernel_regularizer=l2(5e-4)))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.01), metrics=['acc'])

sbalnojan

 print("this simply ignores the connected edges, thus receives a pretty bad test_acc:")
graph_conv_filters = np.eye(A_norm.shape[0])
graph_conv_filters = K.constant(graph_conv_filters)
num_filters = 1

model = Sequential()
model.add(GraphCNN(Y.shape[1], num_filters, graph_conv_filters, input_shape=(X.shape[1],), activation='elu',
                       kernel_regularizer=l2(5e-4)))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.01), metrics=['acc'])

sbalnojan

import os, sys
sys.path.append(os.path.join(os.getcwd(), "keras-deep-graph-learning")) # Adding the submodule to the module search path
sys.path.append(os.path.join(os.getcwd(), "keras-deep-graph-learning/examples")) # Adding the submodule to the module search path
import numpy as np
from examples import utils
from keras.layers import Dense, Activation, Dropout
from keras.models import Model, Sequential
from keras.regularizers import l2
from keras.optimizers import Adam
from keras_dgl.layers import GraphCNN

sbalnojan

train_on_weight= np.array([1,1,0])
print("Now we won't do any fancy preprocessing, just basic training.")

NUM_FILTERS = 1
graph_conv_filters =  A # you may try np.eye(3)
graph_conv_filters = K.constant(graph_conv_filters)

model = Sequential()
model.add(GraphCNN(Y.shape[1], NUM_FILTERS, graph_conv_filters, input_shape=(X.shape[1],), activation='elu', kernel_regularizer=l2(5e-4)))
model.add(Activation('softmax'))

sbalnojan

import os, sys
sys.path.append(os.path.join(os.getcwd(), "keras-deep-graph-learning")) # Adding the submodule to the module search path
sys.path.append(os.path.join(os.getcwd(), "keras-deep-graph-learning/examples")) # Adding the submodule to the module search path
import numpy as np
from keras.layers import Dense, Activation, Dropout
from keras.models import Model, Sequential
from keras.regularizers import l2
from keras.optimizers import Adam
from keras_dgl.layers import GraphCNN
import keras.backend as K

sbalnojan

# ...
# your data preprocessing...
if __name__ == "__main__":

  # ...
  # your data preprocessing...

  s3 = s3fs.S3FileSystem(anon=False)
  df = pd.DataFrame(data={"foo":[0]})
  print(df)