News (February 2022)

• One paper is accepted by ICLR 2022.

• One paper is accepted by ICASSP 2022.

Research

Papers on Generative Models

Zhifeng Kong, Kamalika Chaudhuri. Understanding Instance-based Interpretability of Variational Auto-Encoders. In NeurIPS 2021.
[paper] [code] [video] [blog] [Tag: VAE, Interpretability]

• Zhaoyang Lyu*, Zhifeng Kong*, Xudong Xu, Liang Pan, Dahua Lin. A conditional point diffusion-refinement paradigm for 3d point cloud completion. In ICLR 2022.
[paper] [code] [Tag: DDPM, 3D Point Cloud Completion]

Zhifeng Kong, Wei Ping. On Fast Sampling of Diffusion Probabilistic Models. ArXiv preprint.
[paper] [demo] [code] [Tag: DDPM]

Zhifeng Kong, Wei Ping, Jiaji Huang, Kexin Zhao, Bryan Catanzaro. DiffWave: A Versatile Diffusion Model for Audio Synthesis. In ICLR 2021 (oral).
[paper] [demo] [video] [Tag: DDPM, Speech Synthesis]

Zhifeng Kong, Kamalika Chaudhuri. Universal Approximation of Residual Flows in Maximum Mean Discrepancy. In ICML 2021 INNF Workshop.
[paper] [video] [Tag: Normalizing Flows, Theory]

Zhifeng Kong, Kamalika Chaudhuri. The Expressive Power of a Class of Normalizing Flow Models. In AISTATS 2020.
[paper] [slides] [blog] [Tag: Normalizing Flows, Theory]

Zhifeng Kong, Kamalika Chaudhuri. The Expressive Power of Planar Flows. In The Workshop on Theory of Deep Learning: Where next?
[paper] [slides] [poster] [video] [Tag: Normalizing Flows, Theory]

Other Papers

Zhifeng Kong, Wei Ping, Ambrish Dantrey, Bryan Catanzaro. Speech Denoising in the Waveform Domain with Self-Attention. In ICASSP 2022.
[paper] [demo] [Tag: Self Attention, Speech Denoising]

• Zhangyu Wang, Lantian Xu, Zhifeng Kong, Weilong Wang, Xuyu Peng, Enyang Zheng. A Geometry-Aware Algorithm to Learn Hierarchical Embeddings in Hyperbolic Space. In ICLR 2021 GTRL Workshop.
[paper] [Tag: Representation Learning]

• Zhaoyang Lyu, Ching-Yun Ko, Zhifeng Kong, Ngai Wong, Dahua Lin, Luca Daniel. Fastened CROWN: Tightened Neural Network Robustness Certificates. In AAAI 2020.
[paper] [Tag: Robustness]

Zhifeng Kong. Convergence Analysis of Training Two-hidden-layer Partially Over-parameterized ReLU Networks vis Gradient Descent. In ICNIP 2020.
[paper] [Tag: Optimization, Theory]

Zhifeng Kong. Convergence Analysis of the Dynamics of a Special Kind of Two-Layered Neural Networks with l1 and l2 Regularization. ArXiv preprint.
[paper] [Tag: Optimization, Theory]

• Xingyu Wan, Qing Zhao, Jinjun Wang, Shunming Deng, Zhifeng Kong. Multi-Object Tracking Using Online Metric Learning with Long Short-Term Memory. In IEEE ICIP 2018, 788-792.
[paper] [Tag: LSTM, Object Tracking]

Blogs and Technical Reports

• Understanding Instance-based Interpretability of Variational Auto-Encoders. [blog]

• The expressive power of a class of normalizing flow models. [blog]

• Generalization theory for GANs. [slides]

• Notes on generalization properties of GAN metrics. [paper]

• Invertible ResNets. [slides]

Zhihu Blogs

UCSD Machine Learning Blogs

Industrial Experiences

• Research Internship at Baidu Research, USA. (Summer 2020)

• Research Internship at NVIDIA Applied Deep Learning Research. (Summer 2021)

Talks and Presentations

• I am giving a poster presentation at NeurIPS (December 2021). [video]

• I am giving an invited talk at NeurIPS DGM workshop (December 2021). [video]

• I am giving a talk at UCSD AI seminar (November 2021).

• I am giving two spotlight presentations at ICML INNF workshop (July 2021).

• I am giving a spotlight presentation at ICML XAI workshop (July 2021).

• I am giving an invited talk at NVIDIA (June 2021).

• I am giving an invited talk at AI-Time, an AI community (June 2021).

• I am giving an oral presentation at ICLR 2021 (May 2021). [video]

• I am giving invited talks at TechBeat, an AI community (May 2021). [video]

• I am giving a spotlight presentation at AAAI 2020 (Feb 2020).

• I am giving a spotlight presentation at the IAS deep learning theory workshop (Oct 2019). [video]

Contact me!

Email: z4kong at eng dot ucsd dot edu

WeChat: kong-zhi-feng

Twitter: @ZhifengKong

LinkedIn: Zhifeng Kong

Instagram: zhifeng.kong.3

Zhihu: Zhifeng