• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

"Hard wired" Neural Networks?

krunk7

Member
If I understand correctly, neural networking is now done on a software level. My question: Is there a physical material that accumulates charge until it reaches a critical point before discharging so as to simulate the weighted sums of neuron activity? If so, is there work being done to "hard wire" a neuronal approach to processing? Where can I get information on that?
 
I don't know specifically if it's being done, but I expect it is. It wouldn't be that hard to create a neuron in hardware. A capacitor stores charge, with voltage increasing with increasing charge. A comparator could decide when enough charge has accumulated, then discharge the cap and signal other neurons. The challenge would come in the adaptability. Neural nets, I think, are capable of learning and adapting by changing paths, or re-weighting paths. This type of thing would be possible in hardware, but hard.
 
I expect it could be done but neural nets aren't particularly useful. And, you're neural net would be hardwired so it would have no neuralplasticity which would be really bad.
 
Originally posted by: zephyrprime
I expect it could be done but neural nets aren't particularly useful. And, you're neural net would be hardwired so it would have no neuralplasticity which would be really bad.

Do what sgtroyer said, in an FPGA (one capable of programming itself while running).
 
Its been done before. Mainly more for curiosity sake than for any other reson IIRC. Chips have become so fast nowadays and neural nets still arent ready for mass market so theres no demand for hardware implmentations.
 
Back
Top