I think I have developed a logic circuit which by using combinational logic and flip flops learns to perform the XNOR logic between 2 bits.It is a kind of state machine. Suppose we built a computer made of blocks which use a built in neural network (such as mine logic circuit).Wouldnt that save a lot of memory and let us reach us new heights regarding the ability of the neural networks to solve real life problems?
5
-
2$\begingroup$ Computers are made of building blocks such as your logic circuit. And gates like this can be realized as neural networks, but a neural network that computes XNOR isn't going to magically solve real life problems - it'll just compute XNOR, better or worse. I'm not sure what you're proposing. $\endgroup$Maciej Stachowski– Maciej Stachowski2023-10-13 12:18:27 +00:00Commented Oct 13, 2023 at 12:18
-
$\begingroup$ Instead of programmatically creating artificial networks we can implement the building blocks of the hardware to be a neural network on its own.So that we would save memory and increase processing speed. $\endgroup$Cerise– Cerise2023-10-13 12:26:12 +00:00Commented Oct 13, 2023 at 12:26
-
$\begingroup$ There are chips/cores that provide hardware acceleration of neural networks such as Apple's Neural Engine and NVidia's solutions. But neural networks are still orders of magnitude more complex than simple deterministic Boolean logic - your XNOR realized as a NN would require storing 9 floating-point weights and circuits to perform multiplication and addition, while it only requires 16 transistors when realized deterministically. NNs are good for problems where we don't have a deterministic algorithm. (For girlfriend woes, I recommend Interpersonal Skills SE). $\endgroup$Maciej Stachowski– Maciej Stachowski2023-10-13 12:38:52 +00:00Commented Oct 13, 2023 at 12:38
-
$\begingroup$ Is your logic circuit truly a neural network? Can you teach it to do other things than XNOR? Is it possible you're reinventing an FPGA? $\endgroup$Stack Exchange Broke The Law– Stack Exchange Broke The Law2023-10-17 13:36:01 +00:00Commented Oct 17, 2023 at 13:36
-
$\begingroup$ A FPGA performs any operation based on a deterministic algorithm , which is not what my circuit does. $\endgroup$Cerise– Cerise2023-10-17 14:17:17 +00:00Commented Oct 17, 2023 at 14:17
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
No.
You can design a XNOR logic gate/circuit using standard CMOS transistors (typical transistor-level integrated circuit design) or basic logic gates (i.e., inverters, AND gates, and OR gates). You do not need a flip-flop.
Artificial neural networks perform pattern classification. Certain deep learning architectures, based on generative adversarial networks (GANs), can be used for generative AI. GANs are based on their ability to perform pattern classification.
General-purpose processors, like RISC-V, Intel, and ARM processors, are able to do a lot more computation than artificial neural networks, including deep learning approaches.
