Congratulations to the team at Ayasdi who just published an article in Nature Scientific Reports! It takes a lot of work to get an article published in a peer reviewed journal, so we are really proud of our team who put in the hard work. Because we are changing how people get value from their data, we feel it’s important to communicate to the scientific community about the advances we are making. And besides, publishing papers in peer reviewed journals keeps us scientifically sharp.
The article explains how a new field of mathematics called Topological Data Analysis (TDA) finds shape in complex data and how it is implemented as a software platform. The shape of data shows meaningful properties that explain what groups exist in your data — no matter how complex your data might be. So without having to ask any questions, our networks (please look at the figures) show you how your data is related, so you can get to results faster.
The focus of the article is of topological methods revealing data relationships at a finer scale than commonly used tools, such as principal component analysis or clustering techniques. The article covers three examples of how shape of data reveals insights into three very different datasets: gene expression of tumors, voting behavior in the House of Representatives, and even player positions in the NBA!
Let’s break down one political example in light of our recent presidential election. These data are comprised of voting patterns in the House of Representatives. The data reveals clusters of representatives with different voting tendencies. Automatically and without having much knowledge of the data, we see some major patterns, that voting generally takes place along party lines. No big surprise there! But topology let’s us take a deeper dive into how data is related. So we can see which groups of representatives, within each party, vote in similar ways. And which groups of representatives are likely to collaborate on bipartisan issues.
Check out the image which shows clusters of representatives. Red nodes are Republican and Blue nodes are Democrats. In 2010, voting was highly fragmented in both parties. In 2009, both parties had a much more cohesive voting pattern. Interestingly, Independent Votes (in green) are tightly associated with certain clusters of representatives, indicating that in given years, Independent voters tend to stick to one group’s voting tendencies!