Talk:Reservoir computing

Untitled

Badly needs citations and factual supporting material. Possible sources for a "Further Reading" section include:

http://www.computer.org/portal/web/news/home/-/blogs/researchers-demonstrate-optoelectronic-reservoir-computing?_33_redirect=%2Fportal%2Fweb%2Fnews%2Fhome%3Fp_p_id%3D33%26p_p_lifecycle%3D0%26p_p_state%3Dnormal%26p_p_mode%3Dview%26p_p_col_id%3Dcolumn-1%26p_p_col_pos%3D1%26p_p_col_count%3D3%26_33_struts_action%3D%252Fblogs%252Fview

http://www.technologyreview.com/blog/arxiv/27383/

http://arxiv.org/abs/1111.7219

http://minds.jacobs-university.de/sites/default/files/uploads/papers/2261_LukoseviciusJaeger09.pdf

http://neonasm.blogspot.com/ — Preceding unsigned comment added by 14.162.161.140 (talk) 01:47, 20 February 2014 (UTC)[reply]

Resist the temptation to plagiarize, please.

Another ML article that is abused by shameless researchers trying to advertise their work. The entire "Research initiatives" section should be deleted. It doesn't even have any content. The first part is advertisement, the rest is just random papers that have nothing to do with research initiatives. 2001:16B8:71B7:D000:CC34:72EF:5C81:776D (talk) 14:50, 29 April 2021 (UTC)[reply]

Wiki Education Foundation-supported course assignment

This article was the subject of a Wiki Education Foundation-supported course assignment, between 13 January 2020 and 1 May 2020. Further details are available on the course page. Student editor(s): Ajduarte, Cpowers9810, Gladdenator.

Above undated message substituted from Template:Dashboard.wikiedu.org assignment by PrimeBOT (talk) 03:05, 18 January 2022 (UTC)[reply]

Useless technobabble in the first paragraph

… higher dimensional computational spaces through the dynamics of a fixed, non-linear system …

That means absolutely nothing. And I'm an experiences programmer, have simulated ANNs in the past, and I know exactly what this means: Nothing.

Which system? You mean several related non-linear equations?
What does fixed mean in this context? That it's all constants and no configuration or other inputs besides the data?
If it is fixed, it is not dynamic. Those words mean the opposite of each other!
"computational space" is ridiculously vague. "All natural numbers" could be a "computational space". Anything could. It means nothing.
"higher-dimensional" just means "more input values". If they are created from fewer, either the values are split, or made-up.

I vote for this literally coming out of a bullshit generator, and somebody having put it there to later brag about how certain pseudo-academics will eat any nonsense if it sounds incomprehensible enough.

— 2A00:20:43:788F:96EF:81F7:A793:68C7 (talk) 10:16, 23 September 2021 (UTC)[reply]

I couldn't understand the article, so I asked ChatGPT instead. This explanation made more sense, but is it correct?

 > let's talk about reservoir computing. how do randomly assigned neurons contribute usefully to the output layer? why not just have a single layer of random inputs?
 
In reservoir computing, the randomly assigned neurons form a "reservoir" that provides a high-dimensional, nonlinear transformation of the input. This reservoir captures temporal dynamics and complex patterns in the data. The output layer then performs a linear combination of these transformed inputs to make predictions.

A single layer of random inputs wouldn't capture the temporal dynamics and complex patterns as effectively. The reservoir's high dimensionality and nonlinearity are key to its performance.  — Preceding unsigned comment added by 135.180.194.177 (talk) 00:32, 29 June 2023 (UTC)[reply]