Cracking open the black box of automated machine learning

The interactive tool lets users see and control how the automated model searches work.

Researchers from MIT and elsewhere have developed an interactive tool that, for the first time, lets users see and control how increasingly popular automated machine-learning (AutoML) systems work. Image: Chelsea Turner, MIT
Researchers from MIT and elsewhere have developed an interactive tool that, for the first time, lets users see and control how increasingly popular automated machine-learning (AutoML) systems work. Image: Chelsea Turner, MIT

For the first time, MIT scientists have developed an interactive tool that allows users to see and control how automated machine-learning systems work. Dubbed as ATMSeer, this interactive visualization tool that helps users analyze the searched models and refine the search space.

To ease the difficulty of developing machine learning (ML) models, automated machine learning (AutoML) methods have been proposed. Instead of searching algorithms and tuning hyperparameters manually, AutoML automatically iterates through various machine learning algorithms and optimizes hyperparameters in a predefined search space.

AutoML has received considerable research attention and gained widespread popularity. A plethora of systems for AutoML, such as ATM, SigOpt and Google Cloud AutoML have been developed in recent years. But the systems operate as “black boxes,” meaning their selection techniques are hidden from users. Therefore, users may not trust the results and can find it difficult to tailor the systems to their search needs.

This newly developed tool puts the analyses and control of AutoML methods into users’ hands. It takes as input an AutoML system, a dataset, and some information about a user’s task. Then, it visualizes the search process in a user-friendly interface, which presents in-depth information on the models’ performance.

The tool, ATMSeer, generates a user-friendly interface that shows in-depth information about a chosen models’ performance, as well as the selection of algorithms and parameters that can all be adjusted.  Image: Courtesy of the researchers
The tool, ATMSeer, generates a user-friendly interface that shows in-depth information about a chosen models’ performance, as well as the selection of algorithms and parameters that can all be adjusted.
Image: Courtesy of the researchers

Co-author Kalyan Veeramachaneni, a principal research scientist in the MIT Laboratory for Information and Decision Systems (LIDS), who leads the Data to AI group said, “We let users pick and see how the AutoML systems works. We let users pick and see how the AutoML systems work.”

In studies with science graduate students, scientists found that 85 percent of participants who used ATMSeer were confident in the models selected by the system.

Micah Smith, a graduate student in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in LIDS said, “Micah Smith, a graduate student in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in LIDS.”

Lead author Qianwen Wang of HKUST said, “Data visualization is an effective approach toward better collaboration between humans and machines. ATMSeer exemplifies this idea. ATMSeer will mostly benefit machine-learning practitioners, regardless of their domain, [who] have a certain level of expertise. It can relieve the pain of manually selecting machine-learning algorithms and tuning hyperparameters.”

Auto-Tuned Models (ATM) is the heart of this new tool that catalogs all search results as it tries to fit models to data. It takes as input any dataset and an encoded prediction task. It randomly selects an algorithm class — such as neural networks, decision trees, random forest, and logistic regression — and the model’s hyperparameters, such as the size of a decision tree or the number of neural network layers.

Then, the system runs the model against the dataset, iteratively tunes the hyperparameters, and measures performance. It uses what it has learned about that model’s performance to select another model, and so on. In the end, the system outputs several top-performing models for a task.

The trick is that each model can essentially be treated as one data point with a few variables: algorithm, hyperparameters, and performance. Building on that work, the researchers designed a system that plots the data points and variables on designated graphs and charts. From there, they developed a separate technique that also lets them reconfigure that data in real time.

Lead author Qianwen Wang of HKUST said, “Similar visualization tools are tailored toward analyzing only one specific machine-learning model and allow limited customization of the search space. Therefore, they offer limited support for the AutoML process, in which the configurations of many searched models need to be analyzed. In contrast, ATMSeer supports the analysis of machine-learning models generated with various algorithms.”

ATMSeer’s interface consists of three parts. A control panel allows users to upload datasets and an AutoML system, and start or pause the search process. Below that is an overview panel that shows basic statistics — such as the number of algorithms and hyperparameters searched — and a “leaderboard” of top-performing models in descending order.

ATMSeer includes an “AutoML Profiler,” with panels containing in-depth information about the algorithms and hyperparameters, which can all be adjusted. One panel represents all algorithm classes as histograms — a bar chart that shows the distribution of the algorithm’s performance scores, on a scale of 0 to 10, depending on their hyperparameters. A separate panel displays scatter plots that visualize the tradeoffs in performance for different hyperparameters and algorithm classes.

Results indicate three major factors — number of algorithms searched, system runtime, and finding the top-performing model — determined how users customized their AutoML searches. That information can be used to tailor the systems to users.

Co-author Kalyan Veeramachaneni, a principal research scientist in the MIT Laboratory for Information and Decision Systems (LIDS) said, “We are just starting to see the beginning of the different ways people use these systems and make selections. That’s because now that this information is all in one place, and people can see what’s going on behind the scenes and have the power to control it.”