panrep/decoders.py (16 lines): - line 39: # TODO consider other formulations metapath2vec - line 43: # TODO 2 different type of embeddings context and embedding to improve performance - line 235: # TODO Parameterize per node type pairs!! - line 357: # TODO implement regularization - line 364: # TODO move to dataloader so that it is faster - line 392: # TODO filter edges - line 424: # TODO implement regularization - line 447: # TODO move to dataloader so that it is faster - line 507: # TODO move to dataloader so that it is faster - line 645: # TODO maybe represent triplets as three arrays to make it faster. - line 653: # TODO consider other formulations metapath2vec - line 677: # TODO implement regularization - line 684: # TODO move to dataloader so that it is faster - line 712: # TODO filter edges - line 747: # TODO implement regularization - line 1128: # TODO summary per node type or across all node types? for infomax data_handler/imdb/imbd_data_loader.py (13 lines): - line 68: # TODO genre to multihot encoding - line 133: # TODO Check if this title is important or not.. does it exist in other dictionaries - line 152: # TODO decide if these attributes are needed. - line 204: # TODO is it an entry per person and movie or list of persons - line 210: # TODO Possible Bug. How about when only one or 2 directors or writers exists? - line 237: # TODO test data dumping and loading - line 307: # TODO do not pass the whole value key only the subsets... - line 329: # TODO load first to dictionary and then process in parallel for the nlp model ... - line 395: # TODO Check if this title is important or not.. does it exist in other dictionaries - line 440: # TODO is it an entry per person and movie or list of persons - line 457: # TODO should we include job title and character played ? job category, more meaningfull - line 470: # TODO Possible Bug. How about when only one or 2 directors or writers exists? - line 497: # TODO test data dumping and loading data_handler/imdb/imdb_data_loader.py (13 lines): - line 68: # TODO genre to multihot encoding - line 133: # TODO Check if this title is important or not.. does it exist in other dictionaries - line 152: # TODO decide if these attributes are needed. - line 204: # TODO is it an entry per person and movie or list of persons - line 210: # TODO Possible Bug. How about when only one or 2 directors or writers exists? - line 237: # TODO test data dumping and loading - line 307: # TODO do not pass the whole value key only the subsets... - line 329: # TODO load first to dictionary and then process in parallel for the nlp model ... - line 395: # TODO Check if this title is important or not.. does it exist in other dictionaries - line 440: # TODO is it an entry per person and movie or list of persons - line 457: # TODO should we include job title and character played ? job category, more meaningfull - line 470: # TODO Possible Bug. How about when only one or 2 directors or writers exists? - line 497: # TODO test data dumping and loading panrep/classifiers.py (7 lines): - line 111: # TODO different layers may have different number of hidden units current implementation prevents - line 190: # TODO implement regularization - line 271: # TODO maybe represent triplets as three arrays to make it faster. - line 276: # TODO consider other formulations metapath2vec - line 298: # TODO implement regularization - line 305: # TODO move to dataloader so that it is faster - line 361: # TODO implement regularization panrep/load_data.py (7 lines): - line 499: #TODO fix this is wrong had to add all edges in the testign graph - line 823: # TODO possibly filter out again the frequent nonzero columns - line 906: # TODO THIS IS WRONG!!! I have to add the train valid and test - line 1705: TODO add metapaths - line 1772: TODO add motifs - line 1778: TODO add metapaths - line 1885: TODO add metapaths panrep/layers.py (5 lines): - line 120: # TODO check that the masking step works - line 123: # TODO use sum instead of mean - line 125: # TODO check the masked 1 with without mask that returns the same - line 456: # TODO implement this layer - line 537: # TODO sum for the link prediction to not consider the zero messages data_handler/imdb/imdb_data_to_graph.py (4 lines): - line 19: # TODO how to treat missing attributes - line 153: # TODO change to torch - line 162: # TODO conver to torch th.array() - line 164: # TODO check datatype panrep/edge_masking_samling.py (4 lines): - line 35: #values_o = np.random.randint(num_entity_o, size=num_to_generate) TODO add later - line 258: # TODO 1. Make sure the graph is connected (how) ? rgcn does not - line 271: # TODO negative samples - line 321: # TODO negative samples panrep/evaluation.py (3 lines): - line 25: # TODO have to map the etype and ntype to their integer ids. - line 431: # TODO have to map the etype and ntype to their integer ids. - line 1021: # TODO find all zero indices rows and remove. panrep/encoders.py (3 lines): - line 22: # TODO different layers may have different number of hidden units current implementation prevents - line 73: # TODO possibly high complexity here?? - line 182: # TODO possibly high complexity here?? panrep/utils.py (3 lines): - line 285: # TODO filter positive edges - line 293: # TODO (lingfan): implement filtered metrics - line 327: # TODO filter positive edges