DanieleGregori/ArXivExplore | Paclet Repository

DanieleGregori`ArXivExplore`

ArXivClassifyAuthorsNet

ArXivClassifyAuthorsNet [authorL_:{_Entity..},dim_Integer,drop]
	gives a net classifying a list of author entities 'authorL',using as all linear layers' dimension 'dim',with dropout ratio 'drop';

ArXivClassifyAuthorsNet [authorL_:{_Entity..},dim_List,drop]
	gives a net classifying a list of author entities 'authorL', embedding titles in dimension dim[[1]],abstracts in dimension dim[[2]], (optional) introductions in dimension dim[[4]],catenate them and later passing them into a linear layer with dimension dim[[3]], with dropout ratio 'drop';

ArXivClassifyAuthorsNet [authorL_:{_Entity..},dim_List,drop,idL_List]
	gives a net classifying a list of author entities 'authorL',embedding titles in dimension dim[[1]], abstracts in dimension dim[[2]],introductions in dimension dim[[4]] - building introduction vocabulary with only the list of articles 'idL' - catenate them and later passing them into a linear layer with dimension dim[[3]],with dropout ratio 'drop';

ArXivClassifyAuthorsNet [authorL_List,cat,dim_Integer,drop]
	gives a net classifying the 'authorL' list of author strings in the same category 'cat', using as all linear layers' dimension 'dim', with dropout ratio 'drop';

ArXivClassifyAuthorsNet [authorL_List,cat,dim_List,drop]
	gives a net classifying the 'authorL' list of author strings in the same category 'cat',embedding titles in dimension dim[[1]], abstracts in dimension dim[[2]],catenate them and later passing them into a linear layer with dimension dim[[3]], with dropout ratio 'drop';

ArXivClassifyAuthorsNet [authorL_List,catL_List,dim_List,drop]
	gives a net classifying the 'authorL' list of author strings in categories given by 'catL' list,embedding titles in dimension dim[[1]], abstracts in dimension dim[[2]],catenate them and later passing them into a linear layer with dimension dim[[3]], with dropout ratio 'drop'.

Details and Options

▪

The following options can be given:

This simple recurrent neural network for classifications of ArXiv authors uses essentially a pair of LongShortMemoryLayer on titles and abstracts embedded with their own vocabularies, with before them a DropoutLayer for regularization with default drop ratio set at 0.9. Then a NetGraph combines them with CatenateLayer, allow an intermediate LinearLayer and finally output the classification probabilities through SoftmaxLayer.

Examples

(3)

Basic Examples

(1)

In[1]:=

netA1=

ArXivClassifyAuthorsNet



Sharif (gr)

ARXIVAUTHOR

Cardoso (gr)

ARXIVAUTHOR

Lu (het)

ARXIVAUTHOR

Vafa (het)

ARXIVAUTHOR

,{128,384,64},0.5

Out[1]=

Scope

(1)



Options

(1)

