Computational capabilities of recurrent NARX neural networks.

Computational capabilities of recurrent NARX neural networks. Research Abstract Details 

Research Abstract Table of Contents

Jump to the:

Computational capabilities of recurrent NARX neural networks. Abstract Text:

H T Siegelmann,B G Horne,C L Giles,

Recently, fully connected recurrent neural networks have been proven to be computationally rich-at least as powerful as Turing machines. This work focuses on another network which is popular in control applications and has been found to be very effective at learning a variety of problems. These networks are based upon Nonlinear AutoRegressive models with eXogenous Inputs (NARX models), and are therefore called NARX networks. As opposed to other recurrent networks, NARX networks have a limited feedback which comes only from the output neuron rather than from hidden states. They are formalized by y(t)=Psi(u(t-n(u)), ..., u(t-1), u(t), y(t-n(y)), ..., y(t-1)) where u(t) and y(t) represent input and output of the network at time t, n(u) and n(y) are the input and output order, and the function Psi is the mapping performed by a Multilayer Perceptron. We constructively prove that the NARX networks with a finite number of parameters are computationally as strong as fully connected recurrent networks and thus Turing machines. We conclude that in theory one can use the NARX models, rather than conventional recurrent networks without any computational loss even though their feedback is limited. Furthermore, these results raise the issue of what amount of feedback or recurrence is necessary for any network to be Turing equivalent and what restrictions on feedback limit computational power.

Computational capabilities of recurrent NARX neural networks. Publishing Authors By Initials

HT Siegelmann,BG Horne,CL Giles,

For similar abstracts research abstracts see: abstracts research

PUBMED ID PMID:

MEDLINE DATE:

Computational capabilities of recurrent NARX neural networks. Journal Published:

PUBLICATION TYPE: Journal Article

Journal: IEEE transactions on systems, man, and cybernetics

VOLUME: 27

Page Numbers: 208-15

Journal Abbreviation:

ISSN: 1083-4419

DAY: 7

MONTH: 02

YEAR: 1997

Computational capabilities of recurrent NARX neural networks. Information

Number of References:

LANGUAGE: eng

NlmUniqueID: 9890044

Computational capabilities of recurrent NARX neural networks. Keywords Mesh Terms:

KEYWORDS:

MESH TERMS:

Chemical & Substance for Abstract: Computational capabilities of recurrent NARX neural networks. Information

Substance Name:

Registry Number:

Grant and Affiliation Information for Computational capabilities of recurrent NARX neural networks.

AFFILIATION: Fac. of Ind. Eng. & Manage., Technion-Israel Inst. of Technol., Haifa.

Country: United States

AGENCY:

GRANT:

ACRONYM:

MEDLINETA: IEEE Trans Syst Man Cybern B C

REFSOURCE:

DATABASENAME:

ACCESSION NUMBER:

Number Hits: 0

Computational capabilities of recurrent NARX neural networks Related Publications

• A perspective on telmisartan and cardiovascular risk.
• Adverse childhood experiences and childhood autobiographical memory disturbance.
• Adverse childhood experiences and prescribed psychotropic medications in adults.
• Allele-specific regulation of primary cilia function by the von Hippel-Lindau tumor suppressor.
• Assessing adaptive behaviour in the post-acute setting following traumatic brain injury: initial reliability and validity of the Adaptive Behaviour and Community Competency Scale (ABCCS).
• Bayesian inference for stochastic epidemic models with time-inhomogeneous removal rates.
• Blood pressure-the better biomarker: delay in clinical application.
• Brain stem gliomas: Patterns of Care in Victoria from 1998-2000.
• Cloretazine for the treatment of acute myeloid leukemia.
• Computational capabilities of recurrent NARX neural networks.
• Expression and regulation of anti-mullerian hormone in an oviparous species, the hen.
• Financial implications of plans to combat methicillin-resistant Staphylococcus aureus (MRSA) in an orthopaedic department.
• Hormone therapy and breast cancer: what factors modify the association?
• Institution of universal screening for Group B streptococcus (GBS) from a risk management protocol results in reduction of early-onset GBS disease in a tertiary obstetric unit.
• Is cancer risk associated with anger control and negative affect? Findings from a prospective cohort study.
• Learning long-term dependencies in NARX recurrent neural networks.
• Nelarabine activity in acute biphenotypic leukemia.
• Nicotine in blood in relation to smoking.
• Occupational exposure to ionizing and non-ionizing radiation and risk of glioma.
• On the distribution of performance from multiple neural-network trials.
• Phase I/II study of the mammalian target of rapamycin inhibitor everolimus (RAD001) in patients with relapsed or refractory hematologic malignancies.
• RECENT CASES OF DENTAL MALPRACTICE.
• Recurrent ectopic pregnancy.
• The cardiometabolic syndrome and calcium channel blocker combination drugs.
• The enduring effects of abuse and related adverse experiences in childhood : A convergence of evidence from neurobiology and epidemiology.
• The management of primary cutaneous melanoma in Victoria in 1996 and 2000.
• The oxygen effect on radiation-induced chromosome aberrations: breakage-versus-recombination hypotheses.
• The rs743572 common variant in the promoter of CYP17A1 is not associated with prostate cancer risk or circulating hormonal levels.
• The von Hippel-Lindau tumour suppressor interacts with microtubules through kinesin-2.
• Treatment of acute lymphoblastic leukaemia : a new era.