The best intro for people outside of the field of system design! (I'm personally more familiar with semiconductor device.) Thank you so much. I understood things very clearly :)
If you have a network like this, the currents will not simply be added. In fact you have a complex current divider with multiple voltage sources. The output current is the result of the superposition of all current dividers, which depend on the resistor values and it will get more and more complex by increasing the input vector. Furthermore, the memrisors are changing their values by applying a voltage. How is it possible to get a consistent result?
You know how convolutional neural nets can have a dozen hidden layers etc.. Can you have hidden layers, all in analogue before the ADC? Can you have applications which the analogue output is used directly e.g to drive a servo?
The best intro for people outside of the field of system design! (I'm personally more familiar with semiconductor device.) Thank you so much. I understood things very clearly :)
Do you think a focus on VLSI and circuit design would benefit from this ?
If you have a network like this, the currents will not simply be added. In fact you have a complex current divider with multiple voltage sources. The output current is the result of the superposition of all current dividers, which depend on the resistor values and it will get more and more complex by increasing the input vector. Furthermore, the memrisors are changing their values by applying a voltage. How is it possible to get a consistent result?
fWhat if we ground the lines? Won't we then easily be able to add the currents due to superposition?
Thanks for the clarified explanation.
How can a analog input V and memristor conductance be measured in bits ? (as mentioned 16 bit FP)
Do you by any chance have some publicly available code for this architecture?
Why would we need to keep the resolution as opposed to just making a network which is tolerant of error accumulation?
Wow this is really cool
Does DA and AD conversion needed between each layer !?
You know how convolutional neural nets can have a dozen hidden layers etc.. Can you have hidden layers, all in analogue before the ADC? Can you have applications which the analogue output is used directly e.g to drive a servo?
Maybe two problems would arise: noise and saturation of precision.
Sir i want expalin for memristor based hardware accelerator for image compression videos details plzzz sir