Big Data Startup Ayasdi Launches; Machine Learning Platform Combines Computer Science And Topological Data Analysis
This week a new Big Data startup company launched, Ayasdi, co-founded by Stanford mathematics professor Gunner Carlsson and based on his DARPA-funded research in the area of applied topology, with $10+million in Series A funding led by Khosla Ventures and Floodgate.
The technology, dubbed the Insight Discovery platform, is explained to be the “first machine learning platform that combines computer science and a branch of mathematics known as Topological Data Analysis (TDA) that visualizes the entire dataset.” Hundreds of machine learning algorithms, it says, go to work exploring datasets to in minutes automatically discover insights that can’t be determined through query-based or ad hoc approaches.
Says Carlsson in a video describing the technology, which is based on topology, the subfield of mathematics that concerns itself with the study of shape, “Topological data analysis represents a fundamental advance in machine learning. In the near future, machines will help humans organize, simplify and understand their very large and complicated data sets. This partnership between man and machine will have impact for all areas of human endeavor.”
Based on the Insight Discovery platform, the Ayasdi Iris cloud or onsight solution, is query-Free, model-free, and code-free, the company says. It runs through its machine learning algorithms to automatically construct the underlying shape of the data in powerful visualizations, in order to bring to light what might otherwise be missed insights and anomalies for further investigation. Machine learning algorithms including supervised, unsupervised, and semi-supervised learning and statistical tests are tied together using TDA in the platform, which then highlights patterns or anomolies within data as topological networks.
The company sees applications for the technology in verticals ranging from life sciences to oil and gas to financial services and sports. From the company’s blog, you can get a sense of how such applications play out. For example, Pek Lum, who now leads its products and solutions team, explains how it sussed out something new – a unique data shape brought out by the TDA approach — in an old dataset from the Netherlands Cancer Institute (NKI) created in collaboration with Rosetta.
“After diving deeper, this turned out to be the elusive subgroup of ER negative patients with very good survival. TDA also found another group of ER positive patients with perfect survival that were molecularly different from other ER positive patients.” That led her on a journey around the idea of cancer maps, and to join the company to develop the prototype into an application using TDA “to make a difference in health care.”
The Cancer Genome Atlas is featured on the company’s gallery, which aims at developing targeted cancer treatments, culling through data to identify cancer subtypes that respond differently to treatments.