AIM is accelerating knowledge discovery from high-volume data streams by inventing new techniques to help automate the process of creating and validating hypotheses.
In a world where data are continually streaming from distributed and diverse sources—from scientific instruments, to web traffic, to live imagery—making timely discoveries requires computing capabilities that can keep pace with rapidly evolving phenomena. AIM is developing a new analysis paradigm to provide continuous, automated synthesis of new knowledge and to enable measurement systems to be steered in response to emerging knowledge, rebalancing the effort between humans and machines.
Improving How Humans and Machines Interact
We are transforming the feedback loop in joint human and machine reasoning by capturing human background knowledge through new interaction techniques, allowing AIM systems to evaluate candidate hypotheses rapidly and steer models and data collectors in response to evolving knowledge.
Automating Hypothesis Generation and Testing
We are creating automated methods for generating human-useful hypotheses from data, including incremental machine learning from incomplete data and streaming deductive inference to identify and preserve interesting assertions from live data.
Creating New Streaming Analysis Algorithms
We are developing streaming data characterization methods that can identify and tag important features in high rate data streams and exploring the possibility that streaming algorithms can use dynamic sampling strategies to apply more powerful algorithms to increasingly challenging data rates.
- Arendt wins top prize in international data visualization contest. The IEEE Visualization and Graphics Technical Committee Visualization Pioneers Group International Data-Visualization Contest was held recently at IEEE VIS 2015. Dustin Arendt, along with Yanina Levitskaia (University of Washington), received the Overall Excellence award for their poster "Telling the Story: Gender Discrepancies in Engineering and Computer Science." Contest visualizations were based on the first ever public release of data supplied by education readiness assessment organization, ACT.
- Networking, learning, stimulating new ideas. Lisa Bramer recently attended the 10th Annual Women in Machine Learning Workshop in Montreal, Canada.The technical workshop provides an opportunity for females working in the field to network, exchange ideas, and learn from each other.
- MaTEx version 0.2 now available, combats the issue of scale related to cluster computing. Abhinav Vishnu and team have released The Machine Learning Toolkit for Extreme Scale (MaTEx), an open source software that is available across different platforms, targeted for desktops, supercomputers and cloud computing systems. Version 0.2 is available for download.
- Open source software for the extreme scale. Abhinav Vishnu presented "MaTEx: Machine Learning Toolkit on Extreme Scale" at the Supercomputing 2015 Conference in Austin, TX. View the 20-minute presentation on YouTube.
Employment and Collaboration Opportunities
Want to join us? We're hiring! We're looking for more world-class machine learning and human-computer interaction researchers to join our team. Contact us; tell us what you're passionate about and discuss the possibilities.
We're also interested in partnering with individuals and organizations. If you have expertise in the areas of machine learning or human-computer interaction, and are interested in partnership opportunities, then we'd like to talk with you. Let's collaborate.