Deep Learning Machines are Holographic Memories
Carlos E. Perez
387

So the pattern does not have to be perfect for recognition, but an adequate fit will do. This allows “recognition” as life changes forms, and our perceptions of forms, over time. Having a bit of looseness at every level of the system prevents the loss of potentially useful information. We accidentally eliminate a host of possibilities when we think of things in the wrong context, like not recognizing somebody from one part of life, when seen in another part of life, like work-person seen at non-work-party.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.