Nathan Tsoi (4th year PhD student) will be presenting a new paper on binary neural network classification at the 2022 Neural Information Processing Systems (NeurIPS) conference, one of the top conferences in Artificial Intelligence and Machine Learning. The paper is entitled: “Bridging the Gap: Unifying the Training and Evaluation of Neural Network Binary Classifiers”. The authors are Nathan Tsoi, Kate Candon, Deyuan Li, Yofti Milkessa, and Marynel Vázquez.

While neural network binary classifiers are often evaluated on metrics such as Accuracy and F1-Score, they are commonly trained with a cross-entropy objective. How can this training-evaluation gap be addressed? In the paper, the authors propose a unifying approach to training neural network binary classifiers that combines a differentiable approximation of the Heaviside function with a probabilistic view of the typical confusion matrix values using soft sets. Theoretical analyses shows the benefit of using the method to optimize for a given evaluation metric, such as F1-Score, with soft sets. Also, the extensive experiments show the effectiveness of the approach in several domains.

A more detailed explanation of the proposed approach can be found in the project website. The code is also available in github.com. Try it out!

Congratulations to all the authors and thanks to the National Science Foundation for supporting this work.