Visually Analyzing and Steering Zero Shot Learning
This work addresses the challenge for researchers and practitioners in understanding and improving zero-shot learning models, which is incremental as it builds on existing visualization techniques for model analysis.
The authors tackled the problem of diagnosing and understanding mispredictions in zero-shot learning models by proposing a visual analytics system, which helps users analyze and steer these models to improve performance, as demonstrated through usage scenarios.
We propose a visual analytics system to help a user analyze and steer zero-shot learning models. Zero-shot learning has emerged as a viable scenario for categorizing data that consists of no labeled examples, and thus a promising approach to minimize data annotation from humans. However, it is challenging to understand where zero-shot learning fails, the cause of such failures, and how a user can modify the model to prevent such failures. Our visualization system is designed to help users diagnose and understand mispredictions in such models, so that they may gain insight on the behavior of a model when applied to data associated with categories not seen during training. Through usage scenarios, we highlight how our system can help a user improve performance in zero-shot learning.