Novel Deep Learning Architectures: Classification Accuracy Improvement

EasyChair Preprint 1110

Abstract

In  future,  emotion  classification, object  classification  etc  by  machines  will  play  an  important  role.  In  this  research  paper,  we  proposed a  series  connection  of  Convolutional Neural Network (CNN)  and  Auto-Encoder (AE)  for  classification problems.  We  proposed  a  total  of  three  architectures. We applied  these  architectures  for  emotion  classification. Among  the  three  architectures,  two  architectures  are  trained  with  JAFFE ( Japanese Female  Facial  Expressions),  remaining  one  architecture  was  trained  with  Berlin  Database  of  Emotional  Speech. We  attained  better  classification  accuracy  than  earlier  efforts. We expect  that  such  architectures  will  provide  better  classification  accuracy  in  other  applications  also

Keyphrases: Auto-encoders, Classification, Convolutional Neural Networks, emotion

@booklet{EasyChair:1110,
  author    = {Rama Murthy Garimella and Inthiyaz Basha Kattubadi},
  title     = {Novel  Deep  Learning  Architectures:  Classification  Accuracy  Improvement},
  howpublished = {EasyChair Preprint 1110},
  year      = {EasyChair, 2019}}