Borui Zhang

About me: I am a fourth-year Ph.D. student in the i-VisionGroup at the Department of Automation, Tsinghua University, under the guidance of Professor Jiwen Lu. In 2021, I earned my BE degree from the Department of Automation and a second BA degree from the School of Economics and Management at Tsinghua University.

Research: My research interests span across computer vision and deep learning theory. Currently, my work is centered on:

Interest: Beyond my routine scientific endeavors, I have a profound love for travel, using photography 📷 to seize and immortalize the beauty of life's fleeting moments. In the interstices of my work, I find solace in classical music 🎹, with a particular fondness for Chopin's compositions, which offer a serene balm to the soul and a catalyst for my research creativity. I also keep myself physically active with badminton 🏸, table tennis 🏓, and running 🏃, and I am always on the lookout for new and engaging activities to enrich my life. Life is a continuous journey of discovery and exploration.

Email  /  Google Scholar  /  Github  /  Xiaohongshu  /  CV

* indicates equal contribution

This paper analyzes the consistency of existing Shapley value estimators and proposes the simple amortized estimator, SimSHAP. Extensive experiments conducted on tabular and image datasets validate the effectiveness of our SimSHAP, which significantly accelerates the computation of accurate Shapley values.

* indicates equal contribution

To address the ambiguity in attributions caused by different path choices, we introduced the Concentration Principle and developed SAMP, an efficient model-agnostic interpreter. By incorporating the infinitesimal constraint (IC) and momentum strategy (MS), SAMP provides superior interpretations.

This paper proposes Bort, an optimizer for improving model explainability with boundedness and orthogonality constraints on model parameters, derived from the sufficient conditions of model comprehensibility and invertibility.

This paper proposes an attributable visual similarity learning (AVSL) framework, which employs a generalized similarity learning paradigm to represent the similarity between two images with a graph for a more accurate and explainable similarity measure between images.

This paper proposes to adaptively learn an ensemble of features that characterizes an image from different aspects to model both interclass and intraclass distributions and employs a relational module to capture the correlations among each feature.