Post-doc Research Fellow
Southern University of Science and Technology (SUSTech)
luzhichaocn [at] gmail.com

Zhichao Lu is currently a post-doctoral research fellow with the Dept. of Computer Science and Engineering at the Southern University of Science and Technology, Shenzhen, China. He received the B.Sc. and Ph.D degrees in Electrical and Computer Engineering from Michigan State University, USA, in 2014 and 2020. During Ph.D study, he was a member in the COIN Laboratory, under the supervision of Prof. Kalyanmoy Deb, and collaborated closely with Prof. Vishnu Boddeti, Prof. Erik Goodman, and Prof. Wolfgang Banzhaf.

His main research interest is Computational Intelligence based learning, modeling, and optimization, notably evolutionary multi-objective optimization, automated machine learning, and in particular evolutionary neural architecture search.

Recent News
Selected Research
FaPN: Feature-aligned Pyramid Network for Dense Image Prediction
Shihua Huang, Zhichao Lu , Ran Cheng, Cheng He
ICCV, 2021  
arXiv / code

FaPN a simple yet effective top-down pyramidal architecture to generate multi-scale features for dense image prediction. It improves FPN's AP / mIoU by 1.5 - 2.6% on all tasks.

End-to-End Dense Video Captioning with Parallel Decoding
Teng Wang, Ruimao Zhang, Zhichao Lu , Feng Zheng, Ran Cheng, Ping Luo
ICCV, 2021  
arXiv / code

A transformer-based framework for end-to-end dense video captioning with parallel decoding.

Neural Architecture Transfer
Zhichao Lu , Gautam Sreekumar, Erik Goodman, Wolfgang Banzhaf, Kalyanmoy Deb, Vishnu N. Boddeti
TPAMI, 2021  
arXiv / video / code

Improve practical utilities of neural architecture search through many-objective optimization, iterative surrogate modeling, and transfer learning.

NSGANetV2: Evolutionary Multi-Objective Surrogate-Assisted Neural Architecture Search
Zhichao Lu , Kalyanmoy Deb, Erik Goodman, Wolfgang Banzhaf, Vishnu N. Boddeti
ECCV, 2020   (Oral Presentation)
arXiv / video / code

An efficient multi-objective neural architecture search framework. It can be easily integrated with most of search spaces.

MUXConv: Information Multiplexing in Convolutional Neural Networks
Zhichao Lu , Kalyanmoy Deb, Vishnu N. Boddeti
CVPR, 2020  
arXiv / video / code

A novel convolutional layer designed through neural architecture search to improve both parameter and FLOPs efficiency.

NSGA-Net: Neural Architecture Search using Multi-Objective Genetic Algorithm
Zhichao Lu , Ian Whalen, Vishnu N. Boddeti, Yashesh Dhebar, Kalyanmoy Deb, Erik Goodman, Wolfgang Banzhaf
GECCO, 2019   (Oral Presentation, Best Paper Award)
arXiv / video / code / extended abstract (invited by IJCAI '20)

An evolutionary multi-objective search algorithm for evolving DNN architectures automatically.