Can One Deep Model Be Effective in Multiple Domain? a Case Study with Public Datasets

EasyChair Preprint 11163

Abstract

Deep CNN models like VGG-16, Inception-v3 can be effective in multiple domains to some extent, but their effectiveness depends on several factors, including the specific domains involved, the complexity of the tasks within those domains, and the model's architecture and training data. In this paper we performed an empirical study on the effectiveness of a customized CNN model and tested its efficiency on multiple domains like epidemic disease prediction, NLP applications, and Education Technology. Three public datasets are identified from the review of literature of the existing works.  It has been observed that smaller DCNN are more likely to perform diversely in different domains than larger models that are more robust in performance.

Keyphrases: Convolutional Neural Network, Epidemic Disease prediction, Students' Performance Prediction, deep learning, law section prediction

@booklet{EasyChair:11163,
  author    = {Subhra Dutta and Sushil Mandi},
  title     = {Can One Deep Model Be Effective in Multiple Domain? a Case Study with Public Datasets},
  howpublished = {EasyChair Preprint 11163},
  year      = {EasyChair, 2023}}