OSCR

Data-driven reduced modeling of neural dynamics.

Code ↔ Paper

4 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

2 blocks of code here have a permanent identifier (oscr:b:…), listed with the matches below: citable for good, with a page of its own. Select a match and press Cite to copy it. What a block identifier is.

The 4 matches
  1. [1] § Results › A model for context-dependent decision-making ↔ src/rnn_paper/SSMfunctions.py, lines 93 to 102 · score 0.58 · Manifold Fitting Error, trajectories projected, spectral subspace, MFE, RNN, SSM · block oscr:b:6AERGRS80W1TB-Y · Report a mismatch
  2. [2] § Methods › The Vanilla RNN models › Multitasking RNN: memory-pro task ↔ mt demo.ipynb, lines 42 to 84 · score 0.57 · Memory Pro task, memory period, stimulus, vector, sensory · Report a mismatch
  3. [3] § Methods › The theory of spectral submanifolds › Data-driven computations of SSMs ↔ src/rnn_paper/SSMfunctions.py, lines 93 to 102 · score 0.52 · manifold fitting error, lifts, MFE, trajectories, SSM · block oscr:b:6AERGRS80W1TB-Y · Report a mismatch
  4. [4] § Methods › The Vanilla RNN models › Sine-wave generator RNN ↔ swgRNN/utils/UtilsSamplingLocs.py, lines 140 to 171 · score 0.50 · sine wave generator, network, RNN · block oscr:b:JNX7AZVR6Q695-E · Report a mismatch

Paper

Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 289 lines · 11 KB · MIT · 2 matches

  1. import numpy as np
  2. import scipy as sp
  3. import sympy as sym
  4. from IPython.display import display, Math, Latex
  5. from matplotlib import pyplot as plt
  6. from rnn_paper.utils import GenerateSymbols
  7. # Given coefficients and exponents of the polynomial expansion, computes the polynomial expansion
  8. def evaluate_polynomial(C, I, P, symbolic = False ):
  9. # C is the coefficient (dim(SSM) \times m_order)-matrix
  10. # I is the exponent (m_order \times dim(SSM))-matrix
  11. # P is the d-dimensional vector to expand in monomials
  12. if len(np.shape(P))!= 1:
  13. PHI = np.ones((np.shape(I)[0],*np.shape(P)[1:]), dtype = complex)
  14. else:
  15. PHI = np.ones((np.shape(I)[0]),dtype =complex)
  16. if symbolic:
  17. PHI = sym.MutableDenseNDimArray(PHI)
  18. C = np.round(C, 3)
  19. for i, exp in enumerate(I):
  20. for j in range(np.shape(P)[0]):
  21. PHI[i] = PHI[i]*np.power(P[j], exp[j])
  22. if len(np.shape(P))!= 1:
  23. return (np.tensordot(C, PHI, axes = 1))
  24. else:
  25. return np.dot(C, PHI)
  26. # Given coefficients and exponents of the (autonomous) SSM, computes the SSM expansion
  27. def construct_SSM(y, mfldCoeffs, mfldExps, symbolic = False):
  28. if symbolic == True:
  29. return evaluate_polynomial(mfldCoeffs, mfldExps, y, symbolic = symbolic)
  30. return evaluate_polynomial(mfldCoeffs, mfldExps,y)
  31. # Given coefficients and exponents of the (autonomous) ROM, computes the ROM expansion
  32. def ROM0(eta, rdCoeffs, rdExps, symbolic = False):
  33. if symbolic == True:
  34. return evaluate_polynomial(rdCoeffs,rdExps, eta, symbolic = symbolic)
  35. return np.real(evaluate_polynomial(rdCoeffs,rdExps, eta, symbolic = symbolic))
  36. # To integrate with scipy
  37. def ROM(t,eta, rdCoeffs, rdExps):
  38. return evaluate_polynomial(rdCoeffs,rdExps, eta)
  39. def ROM1D(t, eta, rdCoeffs, rdExps, symbolic = False):
  40. rhs = 0
  41. for i in range(rdCoeffs.shape[1]):
  42. rhs+= eta**rdExps[i,0]*rdCoeffs[0,i]
  43. return rhs
  44. def ROM1D0(eta, rdCoeffs, rdExps, symbolic = False):
  45. rhs = 0
  46. for i in range(rdCoeffs.shape[1]):
  47. rhs+= eta**rdExps[i]*rdCoeffs[0,i]
  48. return rhs
  49. # Print the ROM
  50. def printROM(d,rdCoeffs, rdExps):
  51. etas, detas = GenerateSymbols(d)
  52. # Construct ROM symbolic equations
  53. if d ==1:
  54. rhs = ROM1D0(etas,rdCoeffs, rdExps,symbolic=True)
  55. else:
  56. rhs = ROM0(etas,rdCoeffs, rdExps,symbolic=True)
  57. for i in range(d):
  58. display(Math(sym.latex(sym.Eq(detas[i], rhs[i]))))
  59. # Calculate the Normal Mean Trajectory Error
  60. def calculate_NMTE(nTest, DataTestTrunc, RomTraj):
  61. NMTE = 0
  62. for i in range(nTest):
  63. NMTE += np.mean(np.linalg.norm(DataTestTrunc[i,:,:]-RomTraj[i][:], axis = 0), axis = -1)/np.max(DataTestTrunc[i])
  64. return NMTE/nTest
  65. # Calculate the Manifold Fitting Error
  66. def calculate_MFE(nTest, DataTestTrunc, E, mfldCoeffs, mfldExps):
  67. MFE = 0
  68. for i in range(nTest):
  69. # Lift trajectories projected to the spectral subspace E to the SSM
  70. LiftTraj = construct_SSM(np.dot(E.T,DataTestTrunc[i]),mfldCoeffs, mfldExps)
  71. MFE += np.mean(np.linalg.norm(DataTestTrunc[i,:,:]-LiftTraj[:], axis = 0), axis = -1)/np.max(DataTestTrunc[i])
  72. return MFE/nTest
  73. ####### Time-Dependent SSMs (Weak Forcing) #######
  74. # Anchor traj at order 1
  75. def compute_anchorO1(N, d, t0, nTimesteps, evecs, evals, ns, stable = True, tau = 0.01, dt = 1e-3, tfin = 700):
  76. Sevals = np.copy(evals)
  77. if stable == False:
  78. Sevals[:d] = np.zeros((d), dtype= complex)
  79. Uevals = np.copy(evals)
  80. Uevals[d:] = np.zeros((N-d), dtype= complex)
  81. At = np.exp(np.outer(np.arange(0,nTimesteps*dt, dt), (Sevals)))
  82. Gt = np.zeros((nTimesteps, N,N), dtype = complex)
  83. for i in range(nTimesteps):
  84. Gt[i] = np.diag(At[i])
  85. Gt = np.matmul(evecs,np.matmul((Gt),np.linalg.inv(evecs)))
  86. if stable == False:
  87. Atu = np.exp(np.outer(-np.arange(0,nTimesteps*dt, dt), (Uevals)))
  88. Gtu = np.zeros((nTimesteps, N,N), dtype = complex)
  89. for i in range(nTimesteps):
  90. Gtu[i] = np.diag(Atu[i])
  91. Gtu = np.matmul(evecs,np.matmul((Gtu),np.linalg.inv(evecs)))
  92. T, N, M = Gt.shape
  93. _, M2 = ns.T.shape
  94. assert M == M2, "Dimensions of G(t) and f(t) must align."
  95. # Initialize output
  96. h = np.zeros((T, N), dtype = complex)
  97. # Perform the convolution-like operation
  98. for t in range(T):
  99. for s in range(t ):
  100. h[t] += Gt[t - s] @ ns[:,s]*dt/tau
  101. if stable == False:
  102. for s in range(t,tfin):
  103. h[t] -= Gtu[s-t]@ ns[:,s]*dt/tau
  104. return h
  105. def plotParamSSM(m,tangent_space,eta1_vec,all_coeffsP, fixed_pointsP,ps, p_fpt0, p0, expsP,rd_coeffsP, rd_expsP,parname,colors =['purple','orange','darkgreen', 'darkcyan', 'salmon'], fsize = 17, tsize = 13, dpi = 100 ):
  106. # fig = plt.figure(figsize=(15,8))
  107. fig = plt.figure(layout='constrained', figsize=(15, 6), dpi = dpi)
  108. subfigs = fig.subfigures(1, 2, )
  109. #ax1 = fig.add_subplot(2, 2, 1, )
  110. ax2 = subfigs[1].subplots(subplot_kw={"projection": "3d"})
  111. axsLeft = subfigs[0].subplots(len(ps), 2, sharex=True)
  112. for e, epsilon in enumerate(ps):
  113. for pt in fixed_pointsP[e]:
  114. ptC = (np.tensordot(tangent_space.T,np.array(pt).T-p_fpt0 , axes =1))
  115. ptSSM = construct_SSM([ptC, (epsilon-p0)], all_coeffsP, expsP)
  116. #ax2.plot(ptC,(epsilon-p0),(np.array(pt).T-p_fpt0 )[m],'x', color = colors[e])
  117. ax2.plot(ptC,(epsilon-p0),ptSSM[m],'X', color = colors[e], markersize = 8)
  118. axsLeft[e,0].plot(ptC,(np.array(pt).T-p_fpt0 )[m],'X', color = colors[e], markersize = 8)
  119. #axsLeft[e,0].plot(ptC,ptSSM[m],'', color = colors[e])
  120. axsLeft[e,1].plot(ptC,0,'X', color = colors[e], markersize = 8)
  121. # plot the manifold
  122. ssm = construct_SSM([eta1_vec, np.ones(len(eta1_vec))*(epsilon-p0)], all_coeffsP, expsP)
  123. axsLeft[e,0].plot(eta1_vec,ssm[m], '-',color =colors[e], label = r'$%s =%s$'%(parname,epsilon))
  124. ax2.plot(eta1_vec,(epsilon-p0)*np.ones(len(eta1_vec)),ssm[m], '-',color =colors[e], label = r'$%s =%s$'%(parname,epsilon))
  125. eta1_dot = ROM0([eta1_vec, (epsilon-p0)*np.ones(len(eta1_vec))], rd_coeffsP, rd_expsP)
  126. axsLeft[e,1].plot(eta1_vec, eta1_dot[0], label = r'$%s =%s$'%(parname,epsilon), color = colors[e])
  127. axsLeft[e,1].plot(eta1_vec,np.zeros(len(eta1_vec)),'--', color = 'red',)
  128. axsLeft[-1,0].set_xlabel(r'$\eta_1$', fontsize = fsize)
  129. axsLeft[-1,1].set_xlabel(r'$\eta_1$', fontsize = fsize)
  130. axsLeft[e,0].set_ylabel(r'$y_%s$'%(m+1), fontsize = fsize)
  131. axsLeft[e,1].set_ylabel(r'$\dot{\eta}_1$', fontsize = fsize)
  132. axsLeft[e,0].tick_params(axis='both', which='major', labelsize=tsize)
  133. axsLeft[e,1].tick_params(axis='both', which='major', labelsize=tsize)
  134. axsLeft[e,1].legend(fontsize = fsize, loc = 'upper center')
  135. axsLeft[e,0].legend(fontsize = fsize, loc = 'upper center')
  136. axsLeft[e,1].set_ylim(-1,1)
  137. p_vals = np.linspace(ps[0]-0.001,ps[-1]+0.001,100)
  138. ax2.set_xlabel(r'$\eta_1$',fontsize = fsize)
  139. ax2.set_ylabel(r'$%s-%s0$'%(parname, parname),fontsize =fsize)
  140. ax2.set_zlabel(r'$y_%s$'%(m+1), fontsize = fsize)
  141. #eta1_vec = np.linspace(lims[0]+xmargins[0],lims[1]+xmargins[1],100)
  142. ETA, P = np.meshgrid(eta1_vec, p_vals)
  143. ssm = construct_SSM([ETA, P-p0], all_coeffsP, expsP)
  144. ax2.plot_surface(ETA,P-p0,ssm[m],color = 'mintcream', alpha = 0.2, edgecolors='lightgray', lw = 0.01, zorder = 2)
  145. ax2.legend(fontsize = fsize , )
  146. ax2.tick_params(axis='both', which='major', labelsize=tsize )
  147. ax2.view_init(30, 40, )
  148. #ax2.set_yticks(np.linspace(p_vals[0], p_vals[-1],5))
  149. ax2.set_yticks([])
  150. return fig
  151. # Calculate first order coefficients
  152. '''def calculate_H11(N,d,t0, epsilon,evals, evecs, idx, anchor,d2fy2, ui, tsteps, dtInt, dt = 1e-3 ):
  153. P = evecs[:,np.flip(np.argsort(evals))]
  154. P_inv = np.linalg.inv(P)
  155. A1 = np.copy(evals[d:])
  156. A1 = A1-evals[idx]*np.identity(N-d)
  157. ki = 0
  158. summand = np.zeros(( N-d, int(tsteps+t0/dtInt)), dtype= complex)
  159. h11 = np.zeros((N-d,tsteps),dtype= complex)
  160. for i in np.arange(t0,tsteps+t0, 1):# t
  161. kj = 0
  162. for j in np.arange(t0,i, dtInt): # s
  163. G = np.exp(A1*(i-j)*dt)
  164. m11 = np.matmul(P_inv, np.dot(np.dot(d2fy2,anchor[:,int(j)]/epsilon), np.matmul(P,ui)))[d:]
  165. summand[:,kj] = np.matmul(G, m11)*dt*dtInt
  166. kj = kj+1
  167. h11[:,ki] = (np.sum(summand, axis = 1))
  168. ki = ki+1
  169. return h11'''
  170. def calculate_A1(d, evals, eig):
  171. Ak = np.copy(evals[d:])
  172. Ak = Ak-eig
  173. return Ak
  174. def calculate_H11_vectorized(N, d, evals_sorted, P, P_inv, d2fy2, anchor, ui,
  175. nTimesteps, div=1, dt=0.001, tau=0.01,
  176. order=2, h0='quasistatic'):
  177. """h11(t) solving dh/dt = (Lambda[d:] - lambda_1) h + m11(t),
  178. integrated by exponential time differencing. Returns (T, N-d), complex,
  179. i.e. the same convention as before -> caller still does `.T`.
  180. `tau` is unused (it sits inside evals_sorted); kept for signature compatibility."""
  181. if d != 1:
  182. raise NotImplementedError('written for a 1-D master subspace')
  183. step = dt / div # the *actual* grid step, not dt
  184. T = anchor.shape[1]
  185. # m11(t), contracting d2fy2 with the eigenvector first to avoid an (N,N,T) array
  186. m11 = np.einsum('ijk,j,kt->it', d2fy2, np.matmul(P, ui), anchor, optimize=True)
  187. m11 = np.matmul(P_inv, m11)[d:] # (N-d, T)
  188. A1 = evals_sorted[d:] - evals_sorted[0] # = np.diag(A10)
  189. if np.any(np.real(A1) >= 0):
  190. raise ValueError('no spectral gap: Re(lambda_n - lambda_1) >= 0')
  191. # ETD coefficients (A1 diagonal -> everything elementwise)
  192. z = A1 * step
  193. E = np.exp(z)
  194. small = np.abs(z) < 1e-6 # series guard
  195. zs = np.where(small, 1.0, z)
  196. phi1 = np.where(small, 1 + z/2 + z**2/6, (E - 1) / zs)
  197. phi2 = np.where(small, 0.5 + z/6 + z**2/24, (E - 1 - z) / zs**2)
  198. h = np.zeros((T, N - d), dtype=complex)
  199. h[0] = -m11[:, 0] / A1 if h0 == 'quasistatic' else 0.0 # slaved value; 'zero' also fine
  200. b1, b2 = step * phi1, step * phi2
  201. if order == 1: # m11 piecewise constant
  202. for k in range(T - 1):
  203. h[k+1] = E * h[k] + b1 * m11[:, k]
  204. else: # m11 piecewise linear
  205. for k in range(T - 1):
  206. h[k+1] = E * h[k] + b1 * m11[:, k] + b2 * (m11[:, k+1] - m11[:, k])
  207. return h
  208. def construct_SSM1Dt(uv, t, epsilon, *args, surf =True):
  209. # h0i = 0, h10 = 0, h20, h11,
  210. coeff11, coeff20, coeff30 = args
  211. if surf:
  212. u = np.tile(np.expand_dims(uv, 0), (len(t),1))
  213. return np.concatenate(([u],np.multiply(np.tile(np.expand_dims(coeff20, (1,2)),(1,np.shape(u)[0], np.shape(u)[1])),u**2)+ np.multiply(np.tile(np.expand_dims(coeff30, (1,2)),(1,np.shape(u)[0], np.shape(u)[1])),u**3)
  214. + epsilon*np.multiply(np.tile(np.expand_dims(coeff11[:,t],2),(1,1,np.shape(u)[1])),u)), axis =0)
  215. else:
  216. u = uv
  217. return np.concatenate(([u],np.multiply(np.tile(np.expand_dims(coeff20, (1)),(1,np.shape(u)[0])),u**2)+
  218. + np.multiply(coeff11[:,t],u)*epsilon), axis =0)

SSMfunctions.py at commit f670273, under MIT · at the source

Overview

Authors: A. Marraffa1,2,3, R. Krause4,5, V. Mante4,5, G. Haller1
  1. Institute for Mechanical Systems, ETH Zurich,Zurich, Switzerland
  2. VIB Center for AI and Computational Biology, VIB,Leuven, Belgium
  3. Present Address: Department of Electrical Engineering, KU Leuven,Leuven, Belgium
  4. University of Zurich & ETH Zurich,Zurich, Switzerland
  5. Neuroscience Center Zurich, University of Zurich & ETH Zurich,Zurich, Switzerland
Institutions: ETH Zurich (Switzerland); VIB.AI (Belgium); KU Leuven (Belgium); University of Zurich (Switzerland)
Journal: Nature communications, volume 17, issue 1, article 9123
Dates: received 1 October 2025; accepted 15 July 2026; published online 28 July 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1038/s41467-026-75924-7 · PMID 42649198 · PMCID PMC13519027 · OpenAlex W7171348837
Open access: gold, a free copy (OpenAlex)
Status: code verified (how the code is found and verified)
Categories: computational modeling (no new data) (modality), none (in silico) (organism)
Methods: Smoothing, state filtering, decompositions, Spectral & time-frequency
Keywords: Dynamical systems, Network models
Topic: Model Reduction and Neural Networks (Statistical and Nonlinear Physics, Physics and Astronomy), according to OpenAlex
Citations: not cited yet (Europe PMC); 61 references in the paper

Abstract

Neural ordinary differential equations (ODEs) are widely used in neuroscience to model the collective activity of neurons during behavioral tasks. The high dimensionality of their parameter and activity spaces, however, often make it challenging to infer and interpret the fundamental features of their dynamics.

In this study, we employ recent nonlinear dynamical system techniques to uncover the core dynamics of several Neural ODEs used in contemporary neuroscience. Specifically, using a data-driven approach, we identify Spectral Submanifolds (SSMs), i.e., low-dimensional attracting invariant manifolds tangent to the eigenspaces of fixed points. The internal dynamics of SSMs serve as nonlinear models that reduce the dimensionality of the full RNNs by orders of magnitude. Through low-dimensional, SSM-reduced models, we give mathematically precise definitions of line and ring attractors, which are intuitive concepts commonly used to explain decision-making and working memory. This unprecedented level of understanding of Neural ODEs obtained from SSM reduction enables the interpretation of mathematically well-defined and robust structures in neuronal dynamics, leading to predictions about the neural computations underlying behavior.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repositories

Its files are read in the Code ↔ Paper reader above, with 4 matches between paragraphs and lines of code.

haller-group/SSMTool-2.4

License: GPL-3.0
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 1168f5f574cb40c82c5d6112e41b803c92f945cf, 21 July 2023
Languages: MATLAB (2875), C++ (1)
Size: 3,201 files, 2,876 scripts
Software Heritage: not archived
Found in: the text, “The theory of spectral submanifolds”
Holds: README, license file, tests, documentation, 51 notebooks
Not found: CITATION.cff, environment file, continuous integration
Tools: Optimization Toolbox (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
151 files

haller-group/SSMLearn

License: AGPL-3.0
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 305581114f62239b70c1fe44bce69cbf939326ea, 25 September 2026
Languages: MATLAB (344)
Size: 736 files, 344 scripts
Software Heritage: not archived
Found in: the text, “Data-driven computations of SSMs”
Holds: README, license file, documentation, 32 notebooks
Not found: CITATION.cff, environment file, tests, continuous integration
Tools: Optimization Toolbox (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
12 files

alicemarr/RNN-paper

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: f6702734490c90001db73cd7b7900998fb9a87cf, 28 July 2026
Languages: Python (20), Jupyter (4)
Size: 150 files, 24 scripts
Software Heritage: not archived
Found in: “Code availability”
Holds: README, license file, environment (pyproject.toml, requirements.txt, uv.lock), 4 notebooks
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (21 files), Matplotlib (9 files), SciPy (7 files), SymPy (5 files), h5py (3 files), scikit-learn (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
26 files

Zenodo 20623670

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Size: 1 file
Software Heritage: not checked
Found in: “Data availability”
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
At the source:

Code availability

The code is available at this URL https://github.com/alicemarr/RNN-paper under the MIT license.

Reproduced under the paper's license (CC BY), from the paper cited above.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 4 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 183 scripts, each with its path and the digest of its content;
  • 4 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability

The data is available at this URL https://github.com/alicemarr/RNN-paper under the MIT license. 10.5281/zenodo.20623670.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 2, 28 September 2026

  • Funding: added Eidgenössische Technische Hochschule Zürich

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 4 authors, 2 keywords, 53 references.

Cite

This paper

Marraffa, A., Krause, R., Mante, V., & Haller, G. (2026). Data-driven reduced modeling of neural dynamics. Nature communications, 17(1), 9123. https://doi.org/10.1038/s41467-026-75924-7

BibTeX

@article{marraffa2026data,
author = {Marraffa, A. and Krause, R. and Mante, V. and Haller, G.},
title = {{Data-driven reduced modeling of neural dynamics}},
journal = {Nature communications},
year = {2026},
month = jul,
volume = {17},
number = {1},
pages = {9123},
publisher = {Nature Publishing Group},
issn = {2041-1723},
doi = {10.1038/s41467-026-75924-7},
url = {https://doi.org/10.1038/s41467-026-75924-7},
pmid = {42649198},
pmcid = {PMC13519027}
}

RIS

TY - JOUR
AU - Marraffa, A.
AU - Krause, R.
AU - Mante, V.
AU - Haller, G.
TI - Data-driven reduced modeling of neural dynamics
T2 - Nature communications
J2 - Nat Commun
PY - 2026
DA - 2026/07/28
VL - 17
IS - 1
SP - 9123
SN - 2041-1723
PB - Nature Publishing Group
DO - 10.1038/s41467-026-75924-7
UR - https://doi.org/10.1038/s41467-026-75924-7
LA - en
ER -

CSL-JSON

{
"id": "10.1038/s41467-026-75924-7",
"type": "article-journal",
"title": "Data-driven reduced modeling of neural dynamics",
"container-title": "Nature communications",
"author": [
{
"family": "Marraffa",
"given": "A."
},
{
"family": "Krause",
"given": "R."
},
{
"family": "Mante",
"given": "V."
},
{
"family": "Haller",
"given": "G."
}
],
"container-title-short": "Nat Commun",
"volume": "17",
"issue": "1",
"page": "9123",
"DOI": "10.1038/s41467-026-75924-7",
"PMID": "42649198",
"PMCID": "PMC13519027",
"ISSN": "2041-1723",
"publisher": "Nature Publishing Group",
"URL": "https://doi.org/10.1038/s41467-026-75924-7",
"language": "en",
"issued": {
"date-parts": [
[
2026,
7,
28
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1126/sciadv.adl4743 [code]
Inferring context-dependent computations through linear approximations of prefrontal cortex dynamics.
Journal: Science advances
In common: SciPy, 13 references, author V. Mante
[2] doi:10.1371/journal.pcbi.1013081 [code]
Balanced state of networks of winner-take-all units.
Journal: PLoS computational biology
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, computational modeling (no new data), 13 references
[3] doi:10.1371/journal.pcbi.1011852 [code]
Trained recurrent neural networks develop phase-locked limit cycles in a working memory task.
Journal: PLoS computational biology
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, none (in silico), computational modeling (no new data), 12 references
[4] doi:10.1038/s41467-024-54688-y [code]
Dynamic tracking of objects in the macaque dorsomedial frontal cortex.
Journal: Nature communications
In common: h5py, scikit-learn, SciPy, 2 other tools, computational modeling (no new data), 10 references
[5] doi:10.1038/s41586-024-08433-6 [code]
Individual variability of neural computations underlying flexible decisions.
Journal: Nature
In common: Optimization Toolbox, SciPy, Matplotlib, 1 other tool, computational modeling (no new data), 9 references
[6] doi:10.1038/s41593-025-01869-7 [code]
Latent circuit inference from heterogeneous neural responses during cognitive tasks.
Journal: Nature neuroscience
In common: SciPy, Matplotlib, NumPy, 11 references
[7] doi:10.1101/2025.06.03.657632 [code]
High-dimensional neuronal activity from low-dimensional latent dynamics: a solvable model
Journal: bioRxiv (preprint)
In common: 12 references
[8] doi:10.1038/s41592-024-02582-2 [code]
MARBLE: interpretable representations of neural population dynamics using geometric deep learning.
Journal: Nature methods
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, 10 references
[9] doi:10.1371/journal.pcbi.1010784 [code]
One dimensional approximations of neuronal dynamics reveal computational strategy.
Journal: PLoS computational biology
In common: Optimization Toolbox, 10 references
[10] doi:10.7554/elife.93060 [code]
Aligned and oblique dynamics in recurrent neural networks.
Journal: eLife
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, 9 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Add the badge to the code's README

OSCR: the badge of this paper

If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.

Markdown
<!-- oscr-badge:start -->
[![OSCR](https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7.svg)](https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7)
<!-- oscr-badge:end -->
reStructuredText
.. oscr-badge:start

.. image:: https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7.svg
   :target: https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7
   :alt: OSCR

.. oscr-badge:end
HTML
<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7"><img src="https://openscicode.org/badge/doi/10.1038/s41467-026-75924-7.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->

On GitHub, open the README, choose Edit (the pencil), paste the snippet under the title, then "Commit changes".

If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.

Report a mismatch between the code and the paper · Report a code error (signed in)

Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.

Report a reproduction failure (signed in)

Versions of its code: the releases of its code tied to a version of this paper (the preprint, the accepted manuscript, the version of record…), each with the tracing map versioned with it.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.