{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,14]],"date-time":"2026-02-14T10:29:20Z","timestamp":1771064960128,"version":"3.50.1"},"reference-count":40,"publisher":"MIT Press - Journals","issue":"10","license":[{"start":{"date-parts":[[2020,7,15]],"date-time":"2020-07-15T00:00:00Z","timestamp":1594771200000},"content-version":"vor","delay-in-days":1749,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2015,10,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Simulation of neural behavior on digital architectures often requires the solution of ordinary differential equations (ODEs) at each step of the simulation. For some neural models, this is a significant computational burden, so efficiency is important. Accuracy is also relevant because solutions can be sensitive to model parameterization and time step. These issues are emphasized on fixed-point processors like the ARM unit used in the SpiNNaker architecture. Using the Izhikevich neural model as an example, we explore some solution methods, showing how specific techniques can be used to find balanced solutions.<\/jats:p>\n               <jats:p>We have investigated a number of important and related issues, such as introducing explicit solver reduction (ESR) for merging an explicit ODE solver and autonomous ODE into one algebraic formula, with benefits for both accuracy and speed; a simple, efficient mechanism for cancelling the cumulative lag in state variables caused by threshold crossing between time steps; an exact result for the membrane potential of the Izhikevich model with the other state variable held fixed. Parametric variations of the Izhikevich neuron show both similarities and differences in terms of algorithms and arithmetic types that perform well, making an overall best solution challenging to identify, but we show that particular cases can be improved significantly using the techniques described.<\/jats:p>\n               <jats:p>Using a 1 ms simulation time step and 32-bit fixed-point arithmetic to promote real-time performance, one of the second-order Runge-Kutta methods looks to be the best compromise; Midpoint for speed or Trapezoid for accuracy. SpiNNaker offers an unusual combination of low energy use and real-time performance, so some compromises on accuracy might be expected. However, with a careful choice of approach, results comparable to those of general-purpose systems should be possible in many realistic cases.<\/jats:p>","DOI":"10.1162\/neco_a_00772","type":"journal-article","created":{"date-parts":[[2015,8,27]],"date-time":"2015-08-27T15:15:50Z","timestamp":1440688550000},"page":"2148-2182","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":36,"title":["Accuracy and Efficiency in Fixed-Point Neural ODE Solvers"],"prefix":"10.1162","volume":"27","author":[{"given":"Michael","family":"Hopkins","sequence":"first","affiliation":[{"name":"School of Computer Science, APT Group, University of Manchester, Manchester M13 9PL, U.K. michael.hopkins@manchester.ac.uk"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Steve","family":"Furber","sequence":"additional","affiliation":[{"name":"School of Computer Science, APT Group, University of Manchester, Manchester M13 9PL, U.K. steve.furber@manchester.ac.uk"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"281","published-online":{"date-parts":[[2015,10,1]]},"reference":[{"key":"2022050416563946800_R1","unstructured":"Alsaker, C. A.\n           (2009). Solving systems of differential equations\n            using the Taylor series. (Tech. Rep.). South Dakota School of\n            Mines and Technology."},{"key":"2022050416563946800_R2","unstructured":"Babu\u0161ka, I.,\n                Pr\u00e1ger, M.,\n            & Vit\u00e1sek,\n              E. (1966).\n            Numerical processes in differential equations. New\n            York: SNTL\/Wiley\n          Interscience."},{"key":"2022050416563946800_R3","unstructured":"Bashforth, F., &\n                Adams, J.\n              C. (1883). An attempt\n            to test the theories of capillary action by comparing the theoretical and measured forms\n            of drops of fluid with an explanation of the method of integration employed in\n            construction the tables which give the theoretical forms of such\n            drops.Cambridge: Cambridge\n            University Press."},{"key":"2022050416563946800_R4","doi-asserted-by":"crossref","unstructured":"Brette, R., &\n                Guigon,\n            E. (2003).\n            Reliability of spike timing is a general property of spiking model\n            neurons. Neural Computation, 15,\n            279\u2013308.","DOI":"10.1162\/089976603762552924"},{"key":"2022050416563946800_R5","doi-asserted-by":"crossref","unstructured":"Brette, R.,\n                Rudolph,\n            M., Carnevale,\n                T., Hines,\n              M., Beeman,\n                D., Bower,\n                J.,\n                Diesmann,\n              M.,\u2009\u2026\u2009&\n                Destexhe,\n              A. (2007).\n            Simulation of networks of spiking neurons: A review of tools and\n            strategies. J. Comput. Neurosci., 23,\n            349\u2013398.","DOI":"10.1007\/s10827-007-0038-6"},{"key":"2022050416563946800_R6","doi-asserted-by":"crossref","unstructured":"Butcher, J. C.\n           (2003). Numerical methods for ordinary differential\n            equations. New York:\n            Wiley.","DOI":"10.1002\/0470868279"},{"key":"2022050416563946800_R7","doi-asserted-by":"crossref","unstructured":"Chan, R. P. K., &\n                Tsai, A. Y.\n              J. (2010). On\n            explicit two-derivative Runge-Kutta methods. Numerical\n            Algorithms, 53,\n          171\u2013194.","DOI":"10.1007\/s11075-009-9349-1"},{"key":"2022050416563946800_R8","doi-asserted-by":"crossref","unstructured":"Darulova, E.,\n                Kuncak, V.,\n                Saha, I.,\n            & Majumdar,\n              R. (2013).\n            Synthesis of fixed-point programs (Tech. Rep.).\n            Lausanne: Ecole Polytechnique F\u00e9d\u00e9rale de\n            Lausanne.","DOI":"10.1109\/EMSOFT.2013.6658600"},{"key":"2022050416563946800_R9","unstructured":"Enright, W. H.,\n                Higham, D.\n              J., Owren,\n                B., &\n                Sharp, P.\n              W. (1995). A survey of\n            the explicit Runge-Kutta method (Tech. Rep. 94-291).\n            Toronto: University of\n            Toronto."},{"key":"2022050416563946800_R10","unstructured":"Euler, L.\n           (1913). De integratione aequationum\n            differentialium per approximationem. Opera Omina,\n            ser. 1, 11,\n          424\u2013434."},{"key":"2022050416563946800_R11","doi-asserted-by":"crossref","unstructured":"Furber, S. B.,\n                Galluppi,\n              F., Temple,\n                S., &\n                Plana, L.\n              A. (2014). The\n            SpiNNaker project. Proceedings of the IEEE,\n            102, 652\u2013665.","DOI":"10.1109\/JPROC.2014.2304638"},{"key":"2022050416563946800_R12","doi-asserted-by":"crossref","unstructured":"Gao, X.,\n                Bayliss,\n            S., &\n                Constantinides,\n              G. (2013).\n            SOAP: Structural optimization of arithmetic expressions for high-level\n            synthesis. In Proceedings of the International Conference on\n            Field-Programmable Technology (pp. 112\u2013119).\n            Piscataway, NJ:\n          IEEE.","DOI":"10.1109\/FPT.2013.6718340"},{"key":"2022050416563946800_R13","unstructured":"Gear, C. W.\n           (1971). Numerical initial value problems in ordinary\n            differential equations. Upper Saddle River, NJ:\n            Prentice Hall."},{"key":"2022050416563946800_R14","doi-asserted-by":"crossref","unstructured":"Gill, S.\n           (1951). A process for the step-by-step\n            integration of differential equations in an automatic digital computing\n            machine. Mathematical Proceedings of the Cambridge Philosophical\n            Society, 47(1),\n            96\u2013108.","DOI":"10.1017\/S0305004100026414"},{"key":"2022050416563946800_R15","unstructured":"Hall, G., &\n                Watt, J.\n            M. (Eds.) (1976). Modern\n            numerical methods for ordinary differential equations.\n            Oxford: Clarendon\n          Press."},{"key":"2022050416563946800_R16","doi-asserted-by":"crossref","unstructured":"Hansel, D.,\n                Mato, G.,\n                Meunier,\n            C., &\n                Neltner,\n            L. (1998).\n            On numerical simulations of integrate-and-fire neural\n            networks. Neural Computation, 10,\n            467\u2013483.","DOI":"10.1162\/089976698300017845"},{"key":"2022050416563946800_R17","unstructured":"Henrici, P.\n           (1962). Discrete variable methods in ordinary\n            differential equations. New York:\n            Wiley."},{"key":"2022050416563946800_R18","unstructured":"Heun, K.\n           (1900). Neue Methoden zur approximativen\n            Integration der Differential-gleichungen einer unabh\u00e4ngigen\n            Ver\u00e4nderlichen. Z. Math. Phys., 45,\n            23\u201338."},{"key":"2022050416563946800_R19","unstructured":"Higham, N. J.\n           (1996). Accuracy and stability of numerical\n            algorithms. Philadelphia:\n            SIAM."},{"key":"2022050416563946800_R20","doi-asserted-by":"crossref","unstructured":"Humphries, M. D., &\n                Gurney,\n            K. (2007).\n            Solution methods for a new class of simple model neurons.\n            Neural Computation, 19,\n            3216\u20133225.","DOI":"10.1162\/neco.2007.19.12.3216"},{"key":"2022050416563946800_R21","unstructured":"ISO\/IEC. (2008).\n            TR18037 Programming languages\u2014C \u2013 Extensions to support embedded\n            processors. (Tech. Rep.). Geneva:\n            International Organization for\n          Standardization."},{"key":"2022050416563946800_R22","doi-asserted-by":"crossref","unstructured":"Izhikevich, E. M.\n           (2003). Simple model of spiking\n            neurons. IEEE Trans. on Neural Networks,\n            14, 1569\u20131572.","DOI":"10.1109\/TNN.2003.820440"},{"key":"2022050416563946800_R23","unstructured":"Jin, X.,\n                Furber, S.\n              B., & Woods,\n                J. V. (2008).\n            Efficient modelling of spiking neural networks on a scalable chip\n            multiprocessor. In Proceedings of the IJCNN 2008 (pp.\n            2812\u20132819). New York:\n            Springer."},{"key":"2022050416563946800_R24","doi-asserted-by":"crossref","unstructured":"Kahan, W.\n           (1965). Further remarks on reducing truncation\n            errors. Comm. ACM, 8,\n            40.","DOI":"10.1145\/363707.363723"},{"key":"2022050416563946800_R25","unstructured":"Kutta, M. W.\n           (1901). Beitrag zur n\u00e4herungsweisen\n            Integration totaler Differential-gleichungen. Z. Math.\n            Phys., 46,\n          435\u2013453."},{"key":"2022050416563946800_R26","unstructured":"Lambert, J. D.\n           (1973). Computational methods in ordinary\n            differential equations. New York:\n            Wiley."},{"key":"2022050416563946800_R27","unstructured":"Lambert, J. D.\n           (1991). Numerical methods for ordinary differential\n            systems: The initial value problem. New York:\n            Wiley."},{"key":"2022050416563946800_R28","doi-asserted-by":"crossref","unstructured":"Mainen, Z. F., &\n                Sejnowski, T.\n              J. (1995).\n            Reliability of spike timing in neocortical neurons.\n            Science, 268(5216),\n            1503\u20131506.","DOI":"10.1126\/science.7770778"},{"key":"2022050416563946800_R29","doi-asserted-by":"crossref","unstructured":"Martel, M.\n           (2009). Enhancing the implementation of\n            mathematical formulas for fixed-point and floating-point arithmetics.\n            Formal Methods in System Design, 35,\n            265\u2013278.","DOI":"10.1007\/s10703-009-0068-y"},{"key":"2022050416563946800_R30","doi-asserted-by":"crossref","unstructured":"Mitsui, T.\n           (1982). Runge-Kutta type integration formulas\n            including the evaluation of the second derivative: Part 1. Publ.\n            RIMS, Kyoto Univ., 18,\n          325\u2013365.","DOI":"10.2977\/prims\/1195184026"},{"key":"2022050416563946800_R31","doi-asserted-by":"crossref","unstructured":"Ono, H., &\n                Yoshida,\n            T. (2004).\n            Two-stage explicit Runge-Kutta type methods using\n            derivatives. Japan J. Indust. Appl. Math.,\n            21, 361\u2013374.","DOI":"10.1007\/BF03167588"},{"key":"2022050416563946800_R32","unstructured":"Press, W. H.,\n                Teukolsky, S.\n              A., Vetterling,\n                W. T., &\n                Flannery, B.\n              P. (1992). Numerical\n            recipes in C (2nd ed.). Cambridge:\n            Cambridge University Press."},{"key":"2022050416563946800_R33","doi-asserted-by":"crossref","unstructured":"Runge, C.D.T.\n           (1895). \u00dcber die numerische Aufl\u00f6sung von\n            Differentialgleichungen. Math. Ann.,\n            46, 167\u2013178.","DOI":"10.1007\/BF01446807"},{"key":"2022050416563946800_R34","doi-asserted-by":"crossref","unstructured":"Shintani, H.\n           (1972). On explicit one-step methods utilizing\n            the second derivative. Hirsohima Math. J.,\n            2, 353\u2013368.","DOI":"10.32917\/hmj\/1206137626"},{"key":"2022050416563946800_R35","doi-asserted-by":"crossref","unstructured":"Stewart, R. D., &\n                Bair,\n            W. (2009).\n            Spiking neural network simulation: Numerical integration with the\n            Parker-Sochacki method. J. Comput. Neurosci.,\n            27, 115\u2013133.","DOI":"10.1007\/s10827-008-0131-5"},{"key":"2022050416563946800_R36","doi-asserted-by":"crossref","unstructured":"Stimberg, M.,\n                Goodman, D. F.\n              M., Benichoux,\n                V., &\n                Brette,\n            R. (2014).\n            Equation-oriented specification of neural models for\n            simulations. Frontiers in Neuroinformatics,\n            8(6).","DOI":"10.3389\/fninf.2014.00006"},{"key":"2022050416563946800_R37","doi-asserted-by":"crossref","unstructured":"Touboul, J.\n           (2009). Importance of the cutoff value in the\n            quadratic adaptive integrate-and-fire model. Neural\n            Computation, 21,\n          2114\u20132122.","DOI":"10.1162\/neco.2009.09-08-853"},{"key":"2022050416563946800_R38","doi-asserted-by":"crossref","unstructured":"Touboul, J.\n           (2010). On the simulation of nonlinear\n            bidimensional spiking neuron models. Neural\n          Computation, 23,\n          1704\u20131742.","DOI":"10.1162\/NECO_a_00141"},{"key":"2022050416563946800_R39","doi-asserted-by":"crossref","unstructured":"Vit\u00e1sek, E.\n           (1969). The numerical stability in solution of\n            differential equations. Conf. on Numerical Solution of\n            Differential Equations, 109,\n            87\u2013111.","DOI":"10.1007\/BFb0060017"},{"key":"2022050416563946800_R40","unstructured":"Wolfram Research\n            (2014). Mathematica (version 10.0 ed.).\n            Champaign, IL: Wolfram\n            Research."}],"container-title":["Neural Computation"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/27\/10\/2148\/2019621\/neco_a_00772.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/27\/10\/2148\/2019621\/neco_a_00772.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,5,4]],"date-time":"2022-05-04T16:57:42Z","timestamp":1651683462000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/neco\/article\/27\/10\/2148\/8085\/Accuracy-and-Efficiency-in-Fixed-Point-Neural-ODE"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2015,10,1]]},"references-count":40,"journal-issue":{"issue":"10","published-online":{"date-parts":[[2015,10,1]]},"published-print":{"date-parts":[[2015,10,1]]}},"URL":"https:\/\/doi.org\/10.1162\/neco_a_00772","relation":{},"ISSN":["0899-7667","1530-888X"],"issn-type":[{"value":"0899-7667","type":"print"},{"value":"1530-888X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2015,10]]},"published":{"date-parts":[[2015,10,1]]}}}