DanieleGregori/ArXivExplore | Paclet Repository

DanieleGregori`ArXivExplore`

ArXivFeaturesAuthorsNet

ArXivFeaturesAuthorsNet [authorL_List]
	returns a feature extraction net,modifying the net classifying 'authorL' list of author entities;

ArXivFeaturesAuthorsNet [authorL_List,m,dim_List,drop,rounds]
	returns a feature extraction net, modifying the net classifying 'authorL' list of author entities, with layers dimension list 'dim', dropout level 'drop', trained with 'm' examples for each author, in 'rounds' training rounds .

Examples

(2)

Basic Examples

(1)

In[1]:=

ArXivFeaturesAuthorsNet



Sharif (gr)

ARXIVAUTHOR

,

Cardoso (gr)

ARXIVAUTHOR

,

Lu (het)

ARXIVAUTHOR

,

Vafa (het)

ARXIVAUTHOR

//Reap

Out[1]=

NetGraph

Number of inputs:	2
Outputport:	vector(size: 32)

Data not saved. Save now

,NetTrainResultsObject

NetTrain Results

summary

batches:

810,

rounds:

30,

time:

1.4min

,

examples/s:

316

data

training examples:

864,

validation examples:

96,

processed examples:

25920,

skipped examples:

0

method

ADAM

optimizer

,

batch size

32,

CPU

round

loss:

1.1×

-2

10

,

error:

0.116%

validation

loss:

3.89×

-1

10

,

error:

15.6%

‹

loss

›

	rounds
loss

training set

validation set

,Accuracy0.943396,F1ScoreSharif (gr)0.957746,Cardoso (gr)0.926829,Lu (het)0.931034,Vafa (het)0.952381,ConfusionMatrixPlot

actual class
	predicted class



Scope

(1)



""