Chinese firm unveils world’s first fully integrated analog chip for one-step matrix solving
-
This post did not contain any content.
-
This post did not contain any content.
One-step part confused a bit . Cannot get it , is it still n^3^ operations but in single step means made simultaneously in single cycle?
-
One-step part confused a bit . Cannot get it , is it still n^3^ operations but in single step means made simultaneously in single cycle?
Would need some more technical info on that. I'm guessing what they mean is that the solution falls out from the physical arrangement. So, when you run certain inputs through it, you end up with their corresponding outputs being computed by the way the circuits reinforce or inhibit each other.
-
This post did not contain any content.
-
The AX100 represents an equation as an electrical circuit and reads the answer once the circuit settles into equilibrium
Huh? Settles into equilibrium? What does that mean?
stabilizes
-
The AX100 represents an equation as an electrical circuit and reads the answer once the circuit settles into equilibrium
Huh? Settles into equilibrium? What does that mean?
once the voltages on inputs match the outputs I presume
-
once the voltages on inputs match the outputs I presume
-
stabilizes
-
It would've been great had the article gone a into a bit more depth, even for laymen. It sounds like a magical chip atm.
More detail would've been useful, but this is in no way magical or even new really. Analog chips have been developed for many decades, and it's well understood technology. It's just novel application of it geared towards neural networks.
-
Would need some more technical info on that. I'm guessing what they mean is that the solution falls out from the physical arrangement. So, when you run certain inputs through it, you end up with their corresponding outputs being computed by the way the circuits reinforce or inhibit each other.
I found an YT Video about it, anyway it's an improvement from a project from many years ago, nothing new.
-
I found an YT Video about it, anyway it's an improvement from a project from many years ago, nothing new.
Stability and biasing aside
-
I found an YT Video about it, anyway it's an improvement from a project from many years ago, nothing new.
It depends on the context. Analog is great when you're dealing with gradients, hence why it's such a great fit for neural network. Meanwhile, digital computing is optimized to represent discrete data. Both are tools that have their own distinct uses.
-
One-step part confused a bit . Cannot get it , is it still n^3^ operations but in single step means made simultaneously in single cycle?
I believe here is a related article: https://www.nature.com/articles/s41928-025-01477-0
Ciao! Sembra che tu sia interessato a questa conversazione, ma non hai ancora un account.
Stanco di dover scorrere gli stessi post a ogni visita? Quando registri un account, tornerai sempre esattamente dove eri rimasto e potrai scegliere di essere avvisato delle nuove risposte (tramite email o notifica push). Potrai anche salvare segnalibri e votare i post per mostrare il tuo apprezzamento agli altri membri della comunità.
Con il tuo contributo, questo post potrebbe essere ancora migliore 💗
Registrati Accedi
Citiverse è un progetto che si basa su NodeBB ed è federato! | Categorie federate | Chat | 📱 Installa web app o APK | 🧡 Donazioni | Privacy Policy