Computability and Complexity of Unconventional Computing Devices

Broersma, Hajo; Stepney, Susan; Wendin, Göran

doi:10.1007/978-3-319-65826-1_11

Cited by 7 publications

(6 citation statements)

References 96 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The computing performed in the representation stage here, however, of processing with the output filter, occurs for each run, and so needs to be included in the overall cost of the computation. Such instantiation and representation costs are one way to 'hide' the true cost of unconventional computation [3], and care must be taken to analyse these fully, in addition to the use of other unconventional computing resources [2].…”

Section: Example: Ar Theory Applied To Reservoir Computingmentioning

confidence: 99%

Co-Designing the Computational Model and the Computing Substrate

Stepney

2019

Unconventional Computation and Natural Computation

Self Cite

View full text Add to dashboard Cite

Given a proposed unconventional computing substrate, we can ask: Does it actually compute? If so, how well does it compute? Can it be made to compute better? Given a proposed unconventional computational model we can ask: How powerful is the model? Can it be implemented in a substrate? How faithfully or efficiently can it be implemented? Given complete freedom in the choice of model and substrate, we can ask: Can we co-design a model and substrate to work well together? Here I propose an approach to posing and answering these questions, building on an existing definition of physical computing and framework for characterising the computing properties of given substrates.

show abstract

Section: Example: Ar Theory Applied To Reservoir Computingmentioning

confidence: 99%

Co-Designing the Computational Model and the Computing Substrate

Stepney

2019

Unconventional Computation and Natural Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…The randomly created parameters in W and W in are not adapted. This latter condition has rendered RC interesting for unconventional hardware realizations of RNN-like learning systems, because in principle it allows one to employ any kind of nonlinearly excitable physical medium to instantiate the "reservoir" (1).…”

Section: Problem Statementmentioning

confidence: 99%

“…Furthermore, the energy hunger of classical digital computing technologies is increasingly becoming problematic, both for global energy management and for deploying ever more computing-intense AI algorithms on battery-powered personal devices. All of this has revived interest in "unconventional" computing research, with regards to non-digital material substrates, architectures and algorithms [1]. Among the wide diversity of approaches to unconventional computing, a leading role is emerging for non-digital implementations of artificial neural networks (ANNs).…”

Section: Introductionmentioning

confidence: 99%

Computing optimal discrete readout weights in reservoir computing is NP-hard

Hadaeghi

Jaeger

2019

Neurocomputing

View full text Add to dashboard Cite

We show NP-hardness of a generalized quadratic programming problem, which we called unconstrained n-ary quadratic programming (UNQP). This problem has recently become practically relevant in the context of novel memristor-based neuromorphic microchip designs, where solving the UNQP is a key operation for on-chip training of the neural network implemented on the chip. UNQP is the problem of finding a vector v ∈ S N which minimizes v T Q v + v T c, where S = {s 1 , . . . , s n } ⊂ Z is a given set of eligible parameters for v, Q ∈ Z N ×N is positive semi-definite, and c ∈ Z N . In memristor-based neuromorphic hardware, S is physically given by a finite (and small) number of possible memristor states. The proof of NP-hardness is by reduction from the unconstrained binary quadratic programming problem, which is a special case of UNQP where S = {0, 1} and which is known to be NP-hard.

show abstract

“…Several criticisms can be applied to this idea: even though these phenomena are non-computable, there are no known quantum-based mechanisms that focus on the achievement of intelligence and consciousness; as mentioned above, there is no need to turn to non-computable phenomena to explain intelligence, as this can be achieved using aleatory algorithms; and although this is still controversial, current studies seem to indicate that quantum computers do not perform non-computable processes. Rather, their processes are exactly such as those of a non-quantum computer but are performed in parallel, achieving a much higher speed (Broersma et al, 2017).…”

Section: Previous Hypothesesmentioning

confidence: 99%

A computational theory of consciousness: qualia and the hard problem

García-Baños

2019

View full text Add to dashboard Cite

Purpose This paper aims to contribute to the formulation of a theory of consciousness based only on computational processes. In this manner, sound computational explanations of qualia and the “hard problem” of consciousness are provided in response to a lack of physical, chemical and psychological explanations. Design/methodology/approach The study analyses the little that can be objectively known about qualia, and proposes a process that imitates the same effects. Then it applies the process to a robot (using a thought experiment) to understand whether this would produce the same sensations as humans experience. Findings A computational explanation of qualia and the “hard problem” of consciousness is possible through computational processes. Research limitations/implications This is a proposal, subject to argumentation and proof. It is a falsifiable theory, meaning that it is possible to test or reject it, as its computational basis allows for a future implementation. Practical implications Subjective feeling emerges as an evolutionary by-product when there are no strong evolutionary pressures on the brain. Qualia do not involve magic. These aspects of consciousness in robots and in organisations are capable of being manufactured; one can choose whether to build robots and organisations with qualia and subjective experience. Originality/value To the best of the author’s knowledge, no other computational interpretation of these aspects of consciousness exists. However, it is compatible with the multiple draft model of Dennett (1991) and the attention schema theory of Webb and Graziano (2015).

show abstract

Computability and Complexity of Unconventional Computing Devices

Cited by 7 publications

References 96 publications

Co-Designing the Computational Model and the Computing Substrate

Co-Designing the Computational Model and the Computing Substrate

Computing optimal discrete readout weights in reservoir computing is NP-hard

A computational theory of consciousness: qualia and the hard problem

Contact Info

Product

Resources

About