PHD Research Proposal in Deep Neural Network

Deep Neural Networks serves as the cornerstone technology for several modern artificial intelligence (AI) applications. With the purpose of modeling the complex non-linear association of data, an artificial neural network (ANN) is developed with the multiple hidden layer structure between the input and output layer is referred to Deep Neural Networks (DNN) [1]. The interpolation of the hidden layer in the network heavily relies on the complexity of the system.
DNN with a number of hidden layers brings the higher-level discriminative information concerning the input data. This deeper feature hierarchy allows the DNNs to accomplish outstanding performance in several tasks [2]. It has effectively averted the most critical and unique information loss in each layer of the network. The remarkable success of DNN in acoustic modeling, image recognition, and speech recognition task intends to explode DNN for the number of applications such as language modeling, playing games, cancer detection, self-driving cars, pattern-recognition tasks, and visual classification problems [3].
DNN show their supremacy in the visual classification despite it still confront the severe challenges when dealing with the noisy voice signal in the speech recognition task owing to its higher dependency of a priori knowledge that degrades its performance. Also, it demands longer training time to tune the sensitive parameter that makes DNN as a slow learning model which is one of the significant challenges in the DNN. DNN model has accomplished greater accuracy with the massive amount of training data.
However, there is a need to enhance the learning ability, especially in visual and speech recognition while adopting with the lower amount of training data. Moreover, DNN model lacks in modeling the uncertainty circumstance owing to its ambiguity, vagueness, and impreciseness. It often faces the issue with the evolving nature of the real-time environment owing to the less flexible and the retraining constraints of the classification system.

Reference:

  • [1] Sun, Shizhao, Wei Chen, Liwei Wang, Xiaoguang Liu, and Tie-Yan Liu, “On the depth of deep neural networks: A theoretical view”, In Thirtieth AAAI Conference on Artificial Intelligence, 2016.

  • [2] Sze, Vivienne, Yu-Hsin Chen, Tien-Ju Yang, and Joel S. Emer, “Efficient processing of deep neural networks: A tutorial and survey”, IEEE Proceedings, Vol.105, No.12, pp.2295-2329, 2017.

  • [3] Liu, Weibo, Zidong Wang, Xiaohui Liu, Nianyin Zeng, Yurong Liu, and Fuad E. Alsaadi, “A survey of deep neural network architectures and their applications”, Neurocomputing, Vol.234, pp.11-26, 2017.

Leave Comment

Your email address will not be published. Required fields are marked *

clear formSubmit