Path: ...!2.eu.feeder.erje.net!feeder.erje.net!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Jeroen Belleman <jeroen@nospam.please>
Newsgroups: sci.electronics.design
Subject: Re: Memristor cross bar arrays for faster AI neural nets and math?
Date: Mon, 18 Mar 2024 17:00:23 +0100
Organization: A noiseless patient Spider
Lines: 44
Message-ID: <ut9oc0$8e4e$1@dont-email.me>
References: <ut8hfj$1glje$1@solani.org> <ut93d2$43ac$1@dont-email.me>
 <ut94l2$1gm2b$1@solani.org> <20240318a@crcomp.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Injection-Date: Mon, 18 Mar 2024 15:58:56 -0000 (UTC)
Injection-Info: dont-email.me; posting-host="c4831b78e5e14b522a2ad8ee287138d5";
	logging-data="276622"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX183VVX3f1Xedgkabio039tJ"
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.13.0
Cancel-Lock: sha1:wsefRjURBCB5ZpiJfLhCK8x6/pw=
Content-Language: en-US
In-Reply-To: <20240318a@crcomp.net>
Bytes: 2792

On 3/18/24 13:37, Don wrote:
> Jan Panteltje wrote:
>> Jeroen Belleman wrote:
>>> Jan Panteltje wrote:
>>>> Source:
>>>> University of Massachusetts Amherst
>>>> Summary:
>>>> A team of engineers has proven that their analog computing device, called a
>>>> memristor, can complete complex, scientific
>>>> computing tasks while bypassing the limitations of digital computing.
>>>>    https://www.sciencedaily.com/releases/2024/03/240314145325.htm
>>>>
>>>> bit like our neural nets...
>>>
>>> I have an issue with calling a memristor a 'computing device'. If
>>
>> If you can do matrix computations with it why not?
>> quote from that llnk:
>>   "When organized into a crossbar array,
>>    such a memristive circuit does analog computing by using physical laws
>>    in a massively parallel fashion, substantially accelerating matrix operation
> ,
>>    the most frequently used but very power-hungry computation in neural networks
>>    "
>>
>>> If you accept that, then so are capacitors and inductors!
>>
>> Well you could store analog info in CMOS too, even in capacitors.
>> Inductors? not so sure, not so easy for a long time?
> 
> Does core memory qualify as inductors?
> 
> Danke,
> 

Memory cores have a very wide hysteresis. You need a fair bit of
current to magnetize them, and when they do, they go straight into
saturation and stay there when the current is removed. You have to
reverse the current to magnetize them the other way and again will
flip the whole way.

That's why they were useful as memory. As inductors, not so much.

Jeroen Belleman