Seminar on Thursday August 31, 2017
Time: 14h00-16h00, on Thursday August 31, 2017.
Place: MSLab, 103-C5, ThuyLoi University, 175 Tay Son, Dong Da, Hanoi, Vietnam.
Title: Deep Learning and Its applications
Speaker: Dr. Nguyen Van Nam, University of Engineering and Technology, Hanoi, Vietnam.
The recently emerged industrial revolution 4.0 promises to radically change the social and economic environments. This revolution is based on intelligent robots which are able to replace human beings in every domains from healthcare (elderly assistant robot), commerce (automatic salesman), finance (robo advisor), transport (autopilot), etc. The intelligent robots can be faster built thanks for advances in machine learning, artificial intelligence (AI), robotics, information and communication technologies (ICT). In Vietnam, the fast growing of ICT in recent year has laid a foundation for the intelligent robots. However, their brain can be made from machine learning, AI of which the applications are still inceptive.
In this talk, we present an introduction to deep learning and its applications such as image recognition, automatic speech recognition (ASR), natural language processing, customer relationship management, recommendation systems. Deep learning relies on artificial neural networks (ANN) which include three layers: input (data attributes), hidden (weights, summation function, transformation function) and output layer (classification results). ANNs provide two learning methods: supervised and unsupervised. In the former case, the back propagation techniques is used to train repeatedly so that the output reaches the known desired value. In the latter case, there is no external interception during training process.
We also present an application of deep learning for customer relationship management in vatgia.com. Given the image of a fashion (clothes, shoes, earrings, etc), the system will output its Vietnamese name and its corresponding prices. The system is trained from a huge amount of data from vatgia.com, uses Inception-v3 (open source) and has an error rate of lower than 3.46%.
Nguyen Van Nam received his Master and Ph.D in University of Lyon 1 in 2012. He also graduated the Hanoi University of Science and Technologies in 2001. His scientific interests include some emerging technologies such as high rate computer networks, cloud computing, big data, machine learning, etc.