HyperTendril: Visual Analytics for User-Driven Hyperparameter Optimization of Deep Neural Networks

Fast forward

Direct link to video on YouTube: https://youtu.be/3nD6kXCL2xI

Keywords

Visual analytics, deep learning, machine learning, automated machine learning, human-centered computing

Abstract

To mitigate the pain of manually tuning hyperparameters of deep neural networks, automated machine learning (AutoML) methods have been developed to search for an optimal set of hyperparameters in large combinatorial search spaces. However, the search results of AutoML methods are significantly affected by initial configurations, and it is a non-trivial task to find a proper configuration for them. Therefore, human intervention via a visual analytic approach bears huge potential in this task. In response, we propose HyperTendril, a web-based visual analytics system that supports user-driven hyperparameter tuning processes in model-agnostic environment. HyperTendril utilizes a novel approach to effectively steering hyperparameter optimization (HyperOpt) through an iterative, interactive tuning procedure that allows users to refine the search spaces and the configuration of AutoML method based on their own insights from given results. Using HyperTendril, users can obtain insights into the complex behaviors of various hyperparameter search algorithms and diagnose their configurations. In addition, HyperTendril supports variable importance analysis to help the users refine their search spaces based on the analysis of relative importance of different hyperparameters and their interaction effects. We present an evaluation focusing on how HyperTendril helps users steer their tuning process via a longitudinal user study based on the analysis of interaction logs and in-depth interviews while we deploy our system in a professional industrial environment.