Bringing Wonder to Data: 5-Minute Interview with Weidong Yang

“It was an aha moment. Integrating our solution with Neo4j allowed us to focus on the critical things we do well, visualization and interactivity,” said Weidong Yang, CEO of Kineviz.

San Francisco-based Kineviz believes that working with data is often more of an art than an engineering skill, and so they use subtle artistic cues to surface complex information in high-dimensional datasets. Kineviz specializes in visualizing and presenting complex data so that users can sense interesting things that are going on in their data and use it as input to determine where to dig in to get quantifiable results.

In this week’s five-minute interview (conducted at GraphConnect 2018 in NYC), we spoke with Weidong Yang, CEO of Kineviz, about his insights into why visualization is key to exploring high-dimensional data and the company’s recent work in integrating their product with Neo4j.

What made you choose Neo4j?

Weidong Yang: At Kineviz, we have learned a lot about how we experience things, how we perceive information, using very subtle cues from artistic exploration. And we bring that artistic understanding and exploration into data visualization.

Believe it or not, working with data is often more of an art than an engineering skill. There’s so much information available that you have to use a lot of intuition before you can figure out what is being quantified.

About six years ago, I was working on another startup. We tried to use social networks to build a trust system for the sharing economy. Today there are a lot of companies that do that. It’s quite successful. But around 2011 and 2012, it was still early.

One of the things that we investigated was using networks to figure out how and in what areas you could trust somebody and how people could be paired to work together or share something.

We did a lot of graph work at the time without actually using a graph database. We built our own graph visualization engines in mostly 2D. And the problem we encountered at the time was that once the network graph grew beyond a certain size, it became very difficult to see. All the structure gets smashed together. The local clusters may be separate, but if they have some connections, they are very likely smashed together.

A few years later, we started trying 3D technology, and we realized that it let you see the information much clearer, making the separation of clusters much easier. At the time, we were still using our own in-memory graph engine.

We did a lot of work, trying to build graph algorithms. Last year at this time, believe it or not, we started thinking about Neo4j. We said, wait a minute, Neo4j seems to have solved all of the hard problems we are trying to solve; let’s try it.

We started integrating our solution with Neo4j. That was the aha moment. Things became so much easier, which allowed us to focus on the critical things we do well, visualization and interactivity, and leave a lot of hard data problems to Neo4j. I think it makes a lot of sense.

Tell me more about graph visualization

Yang: I can offer my view about why Neo4j matters for us, especially from the perspective of my own background. I’m a trained physicist. I used to work with quantum dots.

Graph is a high-dimensional problem because of all the nodes and the interactions among them.

Visualization has always been an important tool for me to get insight into complex data. Once you can see something, it helps you to gain some intuition and direction and then you can work to get quantitative results that you can communicate to people.

I’ve found high-dimensional problems particularly challenging. Take, for example, architectural blueprints. If you are a trained architect, you can look at blueprints and make sense of them. But when other people look at blueprints, they have no clue.

Architecture is a three-dimensional structure, and you are trying to represent it into two-dimensional space. It’s very difficult. It takes a lot of brainpower to reconstruct it.

The same thing happens in a high-dimensional dataset. When we deal with a high-dimensional dataset, if you have to look at a two-dimensional visualization, it becomes very difficult. Having the actual dimensions is not an incremental but a qualitative jump.

What do you think is in store for the future of graphs?

Yang: I think graph technology is very promising and very important. For example, we all know that our experience with knowledge is all essentially graphs, connecting how things relate to each other.

But graph technology as a tool for understanding and for analysis is still I would say at the beginning phase. I think there are a lot of problems that need to be resolved. It’s very easy for a graph to become complex. When you only have a few entities, it’s easy to think about it and to reason on the graph. But when the graph becomes complicated, our brain can’t make sense of it because there are too many dimensions, too many things in play.

I think there’s a lot of research that needs to be done in making graphs accessible, easy for people to understand and useful in all kinds of applications.

What do you think is the strongest aspect of Neo4j?

Yang: I’d like to offer one thought from a scientist’s perspective, working with the model and the data. I think there’s a particular area where Neo4j offers an important solution: when you deal with information in a graph that is very complex. Let’s say you have a few labels, but each label has a lot of entities, and there are many, many properties. This is a traditional data problem mixed with a graph problem. And because of its labeled property graph, Neo4j is very well positioned to address this area.

Anything else you’d like to add?

Yang: Although we operate mostly as a consulting firm, people probably don’t know that Kineviz also has a parallel art nonprofit company that brings dance, interactive technology and visualization together.

Want to share about your Neo4j project in a future 5-Minute Interview? Drop us a line at content@neo4j.com