"Some neural networks take different paths. They go at different speeds. But the striking thing is they're all going the same way," said James Sethna, professor of physics at Cornell.
The team's technique could potentially become a tool to determine which networks are the most effective.
[
add
]
[
|
|
...
]