Cohesion networks, often known as “spillovers” or “sparklines” in the network literature, are highly interactive, allowing researchers to capture and manipulate the behaviour of networked objects.

In a paper published in the journal Science, scientists from the University of Manchester, University of Edinburgh, the University College London and the University Centre for Systems Neuroscience in Oxford describe how they used tensor flow network analysis to find how an object behaves when it crosses the threshold for a sparkline.

The scientists then analysed the data to determine whether a spark line was an important part of the network. 

“We found that there is an important element of the neural network that can detect sparklines in an object,” said lead author Dr Jonathan Wood.

“This is because these lines are not the only thing that the object is responding to.

If the sparkline can be detected, then you can then use this information to predict the behaviour that a spark will have.” 

While the scientists found a few sparklines within the network, the majority of the object’s behaviour was determined by the shape of the spark line, which was not detected. 

A similar technique was used by researchers in 2013 to analyse “spooky” patterns in the behaviour and movement of a number of objects including jellyfish and spiders.

“Spooky” phenomena, such as jellyfish swimming in strange patterns, have become popular as a way to study the behaviour in a network.

Sparklines were first identified as a means to study this phenomenon in a 2009 paper, when a team from the UK’s University of Bristol discovered that jellyfish can emit electrical pulses when they change colour.

A similar phenomenon was discovered in the 2013 paper, which showed that a group of jellyfish in the southern Indian Ocean can change colour when their electrical activity changes.

The authors found that this electrical activity is the only signal that can tell whether an object is moving towards a spark or away from it.

The study also showed that the number of sparklines an object had varied significantly from one experiment to the next, suggesting that the sparklines had a different shape and were less likely to form in a single object.

As well as being a useful method to study objects in networks, the scientists said that it can also be used to learn about the properties of objects and their connections in the real world. 

The study found that in real life, objects in a cluster have a greater chance of forming sparklines when they are travelling towards each other than when they were separated.

“We see this when we look at the patterns of activity of these objects as they move from one object to another, and when we move the objects around in the room, they tend to be clustered together,” said Dr Wood.

 “Our results suggest that if we are able to predict a few details of the behaviour associated with sparklines, we can understand how these clusters are formed and what information is needed to control the behaviour.”### The findings were published in Science.