Pith. sign in

REVIEW 3 major objections 3 minor 1 cited by

Gradient Flow Equations for Deep Linear Neural Networks: A Survey from a Network Perspective

T0 review · 3 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that the entire gradient-flow training of a deep linear network can be rewritten as a single matrix ODE for the network's block-shift adjacency matrix, and that from this representation one can read off stable and unstable

desk verdict The survey is solid, but the headline stability theorem (Thm. 24) is unsupported and contradicted by the paper's own example. read the letter →

arxiv 2511.10362 v1 pith:KQGGXPKK submitted 2025-11-13 cs.LG cs.SYeess.SYmath.DS

Joel Wendin , Claudio Altafini This is my paper · ORCID
classification cs.LGcs.SYeess.SYmath.DS MSC 34D0515A1837C1068T07
keywords gradientflowdeeplinearneuralnetworksadjacencymatrixlosslandscapesaddlepointsquotientspaceisospectralODEconservationlaws
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that the gradient flow of a deep linear network is naturally one polynomial, nilpotent, isospectral matrix ODE written on the adjacency matrix of the network, whose h-th power is the input-output map. In this representation the loss is a quadratic function of that power, and every critical point is labeled by which singular values of the data it has learned. The central claim is that the quotient structure induced by the power map lets one determine stable and unstable submanifolds of every critical point by drawing arcs between critical points and checking monotonicity of the loss, even where the Hessian is singular and silent. If right, this gives a global, non-local map of the loss landscape and explains why gradient descent navigates infinitely many degenerate saddles successfully.

What carries the argument

The key object is the block-shift adjacency matrix A = blk₁(W₁,…,Wₕ), a matrix whose only nonzero blocks sit one diagonal below the main block diagonal, so that the input-output map equals Aʰ. Two derived structures carry the argument: the quotient map φ(A)=Aʰ, whose fibers collect all networks with the same critical value, and the arc A(α) ∈ φ⁻¹((1−α)Φ₁+αΦ₂) connecting two critical points, along which monotonicity of L signals stable and unstable submanifolds. The block-shift SVD alignment and the conserved quantities JQ=C justify restricting to tractable invariant submanifolds such as balanced and decoupled dynamics.

What would settle it

For a small two- or three-layer network, compute the image of the map A ↦ Aʰ and check whether the linear interpolation Φ(α) = (1−α)Φ₁ + αΦ₂ between two critical products stays in that image for all α ∈ [0,1]; exhibit any nested pair for which some α is not the h-th power of any block-shift matrix. That would sever the arc on which Theorem 24's monotonicity argument rests.

Watch

Extended reading notes

Core claim

The central discovery is that the gradient flow of a deep linear network is a matrix ODE Ȧ = Σⱼ (A^{h−j})ᵀ (E−Aʰ)(A^{j−1})ᵀ on the block-shift adjacency matrix A = blk₁(W₁,…,Wₕ), with E encoding the data singular values. Because A is nilpotent, products of layer matrices become powers of A, and the loss becomes ½‖E−Aʰ‖²_F. The critical points are exhausted by binary signatures s ∈ {0,1}^{d_y}: the signature s=1 gives global minima, every other signature gives a saddle point, strict or non-strict depending on the parametrization A = P(A₁+Z)P⁻¹. The paper proves that through the quotient map φ(A)=Aʰ and its fibers φ⁻¹(Φ), one can classify stable and unstable submanifolds of all critical points

Load-bearing premise

The argument's load-bearing premise is that between any two critical points with nested signatures there exists a continuous arc of adjacency matrices whose input-output power interpolates linearly between the two critical values, and that monotonicity of the loss along that arc transfers to the actual training flow; the paper assumes this arc rather than proving it exists.

Editorial extensions

If this is right

  • Every trajectory converges to a single critical point, and the only attractors are global minima; every non-global critical point is a saddle, including non-strict saddles where the Hessian is positive semidefinite.
  • In the quotient space Aʰ there are exactly 2^{d_y} isolated critical points, one per critical value, with critical values given by distinct partial sums of the squared data singular values.
  • For a critical point of signature S, the paper counts 2^{d_y−|S|}−1 unstable submanifolds and 2^{|S|}−1 stable ones, without computing eigenvalues of the Hessian.
  • Under 0-balanced initialization, all layers synchronize their singular values and the critical point is maximally tightened; near-balanced initialization generically yields sequential learning from the largest to the smallest singular value when the scalar channels start equally.
  • Restricting to block-shift diagonal or decoupled dynamics reduces the gradient flow to d_y independent scalar generalized logistic ODEs, reproducing the plateau-and-discovery behavior seen in simulations.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the existence of the arc A(α) could be established in general, the monotonicity criterion would provide a global partial order on critical points and a topological proof of saddle-to-saddle learning, potentially extending beyond quadratic losses where Hessian methods fail.
  • The quotient/fiber picture suggests a testable bridge to matrix sensing and matrix completion: the greedy low-rank principle should be expressible as a statement about which arcs between critical points are realizable in the fiber, not merely which critical points are attractive.
  • A concrete empirical check: in a three-layer near-balanced network, trajectories should approximately follow the monotone arcs of Theorem 24; if they do, the arcs are effective descriptors of learning despite not being actual flow trajectories.
  • The paper's claim that implicit low-rank regularization is confined to near-balanced initialization, rather than universal, invites an experiment sweeping initialization scale and measuring the rank of A along trajectories to delineate where the greedy low-rank story holds.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. The paper surveys the gradient flow dynamics of deep linear neural networks with quadratic loss, reformulating the standard layer-wise equations (9) as a single matrix ODE (12) for the block-shift adjacency matrix A = blk1(W1,...,Wh). Sections 3–5 derive structural properties (nilpotency, conservation laws, Hessian quadratic form, aligned/decoupled/balanced reductions) and review known critical-point and convergence results from [1,10,24,40] in this notation. The claimed original contribution is Theorem 24, which uses monotonicity of the loss along arcs in the fibers of the quotient map φ(A)=A^h to classify stable and unstable submanifolds of critical points, advertised in the abstract as allowing one to determine stable and unstable submanifolds 'even when the Hessian fails to obtain them.'

Significance. The adjacency-matrix formulation is a useful organizing tool: it compresses the multi-layer equations, makes conservation laws transparent (Prop. 6), and simplifies several second-order computations (Props. 20–23). As a survey, the paper collects and reformulates a substantial body of recent results, and the derivations in Sections 3–5 are mostly clean. The claimed stability classification, however, is not established. Theorem 24 infers dynamical invariant manifolds from monotonicity of a Lyapunov function along an arbitrary connecting arc, and the authors explicitly note that the arc need not be a gradient trajectory. The theorem is also contradicted by the paper's own Example 6.2.1. Since this theorem is the main original basis for the abstract's central claim, the significance of the paper as a research contribution is conditional: if Theorem 24 is removed or downgraded to a statement about loss monotonicity, the survey remains valuable but the advertised new tool is lost.

major comments (3)
  1. [§4.6, Theorem 24(2) and following paragraph] The theorem asserts that monotonicity of L(A(α)) along an arc A(α) connecting critical points implies that 'A1 has an unstable submanifold, while A2 has a stable one.' The text immediately after the proof states that these arcs 'cannot be trajectories of the gradient flow' and that 'we do not know how the arcs A(α) relate to the trajectories of (12).' Monotonicity of a Lyapunov function along an arbitrary curve between equilibria does not, by itself, imply the existence of invariant manifolds of the flow. A proof that the arc can be realized by—or is tangent to—actual gradient trajectories is needed. As written the inference is a non sequitur, and it is the load-bearing step for the abstract's claim about determining stable/unstable submanifolds.
  2. [§4.6, Theorem 24(3) vs. §6.2.1] For the single-hidden-node example, dy=1 and the origin has signature S=∅. Item 3 of Theorem 24 predicts 2^0−1=0 stable submanifolds for this critical point. Section 6.2.1, however, explicitly identifies the line w1=−w2 as the stable submanifold of the origin, with trajectories converging to it. Either 'stable submanifold' is being used in a nonstandard, undefined sense, or the count in Theorem 24(3) is wrong. This internal contradiction is decisive for the advertised stability classification.
  3. [§4.6, statement of Theorem 24] The theorem assumes an arc A(α) ∈ φ^{-1}(Φ(α)) with A(0)=A1 and A(1)=A2. No construction or existence proof is supplied, and the parametrization (32)–(33) does not make such a continuous lift with prescribed endpoints evident. Existence must be proved, or explicitly stated as a hypothesis with its limitations analyzed; otherwise the classification applies only to arcs whose existence is itself unverified.
minor comments (3)
  1. [Abstract and §4.3] 'Level sets are unbounded invariant sets of critical points' is imprecise: the full level sets of L contain noncritical points. Only the critical level sets M_c of Prop. 15 are invariant and composed of critical points. Suggest rephrasing.
  2. [§2.1 and §4.2] The symbol S is used both for the indicator set {i : s_i=1} and for the diagonal matrix S=diag(s), sometimes in the same passage (e.g., Prop. 14). A separate symbol for the diagonal matrix would reduce ambiguity.
  3. [§5.4, Eq. (61)] The claim that the graph of ˙z = z^{h-1}(σ_j − z^h) has '3 critical points ±σ_j and 0' is not correct in general. The nonzero critical points are the real roots of z^h=σ_j; there are two of them only when h is even, and they are not ±σ_j unless h=1 or the notation is otherwise defined.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's derivations are explicit algebraic reformulations or external citations; the unsupported arc-based stability claim is a correctness gap, not circularity.

full rationale

The paper's central derivations do not reduce to their inputs. The adjacency-matrix gradient flow (Prop. 4) is obtained by explicit algebra from the standard layerwise equations (9). Critical point classification (Props. 14, 18) is cited from external works [1,24,43] and the Hessian computations are self-contained given those characterizations. No fitted parameters or load-bearing self-citations appear. The original claim in Theorem 24 about stable/unstable submanifolds via arcs is unsupported, and the paper itself notes that the arcs cannot be trajectories of the gradient flow and that their relation to trajectories is unknown. That is a correctness gap, not a circular reduction, because Theorem 24 does not assume the conclusion it derives; it draws a dynamical inference from monotonicity along an arbitrary curve. The stable-manifold inconsistency with Sec. 6.2.1 is also a correctness issue. Thus score 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard assumptions for deep linear networks, on cited structural theorems, and on an unproven arc-existence assertion in Theorem 24. There are no fitted free parameters or invented entities.

assumptions (4)
  • domain assumption Assumption 1: overdetermined regression, X and Y full row rank, and hidden layers at least as wide as output (d_i ≥ d_y).
    Used to reformulate the loss as (5), guarantee surjectivity of φ (filling), and apply prior critical-point results.
  • domain assumption Assumption 2: nonzero distinct singular values and distinct partial sums of squares.
    Ensures the 2^dy critical values are distinct, making the level sets Mc disjoint in Prop. 15 and Prop. 17.
  • domain assumption Cited structural results from [1], [10], [24], [43], [40]: no bad local minima, all non-global critical points are saddles, parametrization of critical points, strict/non-strict classification.
    The survey states these 'without proofs' in Section 4.2 and uses them to build the adjacency-matrix picture.
  • ad hoc to paper Existence of arcs A(α) in the fiber φ⁻¹(Φ(α)) through two prescribed critical points A₁, A₂.
    The paper asserts existence in Section 4.6 without proof; it is needed for Theorem 24's submanifold counting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gradient Flow Equations for Deep Linear Neural Networks: A Survey from a Network Perspective." pith.science (2026). https://pith.science/paper/KQGGXPKK

@misc{pith2026251110362,
  author       = {Pith},
  title        = {Pith review of: Gradient Flow Equations for Deep Linear Neural Networks: A Survey from a Network Perspective},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KQGGXPKK}},
  note         = {Machine review of arXiv:2511.10362}
}
read the original abstract

The paper surveys recent progresses in understanding the dynamics and loss landscape of the gradient flow equations associated to deep linear neural networks, i.e., the gradient descent training dynamics (in the limit when the step size goes to 0) of deep neural networks missing the activation functions and subject to quadratic loss functions. When formulated in terms of the adjacency matrix of the neural network, as we do in the paper, these gradient flow equations form a class of converging matrix ODEs which is nilpotent, polynomial, isospectral, and with conservation laws. The loss landscape is described in detail. It is characterized by infinitely many global minima and saddle points, both strict and nonstrict, but lacks local minima and maxima. The loss function itself is a positive semidefinite Lyapunov function for the gradient flow, and its level sets are unbounded invariant sets of critical points, with critical values that correspond to the amount of singular values of the input-output data learnt by the gradient along a certain trajectory. The adjacency matrix representation we use in the paper allows to highlight the existence of a quotient space structure in which each critical value of the loss function is represented only once, while all other critical points with the same critical value belong to the fiber associated to the quotient space. It also allows to easily determine stable and unstable submanifolds at the saddle points, even when the Hessian fails to obtain them.

Figures

Figures reproduced from arXiv: 2511.10362 by the authors.

Figure 1
Figure 1. Example in Section 5.4. (a): Graph (sj , s˙j ) of the scalar equation (61). Green dots are stable critical points. The origin (in red) is unstable. The lower panel is a zoomed-in version of the upper panel. (b): Singular values of any layer along two gradient flow trajectories. On the upper panel sj (0) = sk(0) and the learning is sequential. On the lower panel sj (0) ̸= sk(0) and the learning is not sequential. 6 E… view at source ↗
Figure 2
Figure 2. Example in Section 6.1. (a): Case with A(0) near the origin. (b): Case with A(0) away from the origin. Convergence is to a global minimum in both panels. (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p033_2.png] view at source ↗
Figure 3
Figure 3. Example of Section 6.1. Singular values and vectors for δ-balanced trajectories, δ = 0.04. (a): Singular values of Σ have relatively close spacing, σj − σj+1 = 1.5, ∀j. The singular vectors approximately align. (b): Singular values of Σ are pairwise very close, σj − σj+1 = 0.1, j = 1, 3. The singular vectors do not align. 33 [PITH_FULL_IMAGE:figures/full_fig_p033_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Example in Section 6.1. Arcs L(A(α)) for the various cases of Theorem 24. equilibria, we get that each trajectory not passing through the origin has exactly one equilibrium point. In fact, the algebraic system ( σ − w1w2 = 0 w 2 1 − w 2 2 = c (63) leads to w 4 1 − cw2 …
Figure 5
Figure 5. Figure 5: Example of 1 layer with 1 hidden node. Green dots are global minima. The origin in red is [PITH_FULL_IMAGE:figures/full_fig_p035_5.png]
Figure 6
Figure 6. Figure 6: Example of 2 hidden layers with one node each (3D phase space). Colored points represent [PITH_FULL_IMAGE:figures/full_fig_p036_6.png]
Figure 7
Figure 7. Figure 7: Example of 1 hidden layer with 2 nodes. (a): 3D slices of the 4D parameter space. Green [PITH_FULL_IMAGE:figures/full_fig_p037_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Statistical Inference on Gradient Flows

    math.ST 2026-05 unverdicted novelty 7.0 of 10

    Proves uniform CLT for gradient flows in ERM and constructs an algorithm-aware, inversion-free covariance estimator for asymptotically valid time-uniform confidence intervals.

Reference graph

Works this paper leans on

45 extracted references · 3 linked inside Pith · cited by 1 Pith paper

  1. [1]

    E. M. Achour, F. Malgouyres, and S. Gerchinovitz. The loss landscape of deep linear neural networks: a second-order analysis.J. Mach. Learn. Res., 25(242):1–76, 2024

    2024

  2. [2]

    Arora, N

    S. Arora, N. Cohen, N. Golowich, and W. Hu. A convergence analysis of gradient descent for deep linear neural networks. InProceedings of the International Conference on Learning Representa- tions, 2019

    2019

  3. [3]

    Arora, N

    S. Arora, N. Cohen, and E. Hazan. On the optimization of deep networks: Implicit acceleration by overparameterization. InProceedings of the International Conference on Machine Learning, pages 244–253. PMLR, 2018

    2018

  4. [4]

    Arora, N

    S. Arora, N. Cohen, W. Hu, and Y. Luo. Implicit regularization in deep matrix factorization. In Advances in Neural Information Processing Systems, 2019. 43

    2019

  5. [5]

    B. Bah, H. Rauhut, U. Terstiege, and M. Westdickenberg. Learning deep linear neural networks: Riemannian gradient flows and convergence to global minimizers.Information and Inference: A Journal of the IMA, 11(1):307–353, 2022

    2022

  6. [6]

    Baldi and K

    P. Baldi and K. Hornik. Neural networks and principal component analysis: Learning from examples without local minima.Neural Netw., 2(1):53–58, 1989

    1989

  7. [7]

    M. A. Belabbas. On implicit regularization: Morse functions and applications to matrix factor- ization.arXiv preprint arXiv:2001.04264, 2020

    arXiv 2001

  8. [8]

    Blum and R

    A. Blum and R. Rivest. Training a 3-node neural network is np-complete. InAdvances in Neural Information Processing Systems, 1988

    1988

Show all 45 references
  1. [9]

    A. A. Blumberg. Logistic growth rate functions.J. Theor. Biol., 21(1):42–44, 1968

    1968

  2. [10]

    Chitour, Z

    Y. Chitour, Z. Liao, and R. Couillet. A geometric approach of gradient descent algorithms in linear neural networks.Math. Control Relat. Fields, 13(3):918–945, 2023

    2023

  3. [11]

    Chizat, M

    L. Chizat, M. Colombo, X. Fern´ andez-Real, and A. Figalli. Infinite-width limit of deep linear neural networks.Comm. Pure Appl. Math., 77(10):3958–4007, 2024

    2024

  4. [12]

    Cohen and N

    N. Cohen and N. Razin. Lecture notes on linear neural networks: A tale of optimization and generalization in deep learning.arXiv preprint arXiv:2408.13767, 2024

    2024 arXiv

  5. [13]

    B. L. De Moor. On the structure and geometry of the product singular value decomposition. Linear Algebra Appl., 168:95–136, 1992

    1992

  6. [14]

    C. C. Domin´ e, L. Braun, J. E. Fitzgerald, and A. M. Saxe. Exact learning dynamics of deep linear networks with prior knowledge.J. Stat. Mech.: Theory Exp., 2023(11):114004, 2023

    2023

  7. [15]

    S. S. Du, W. Hu, and J. D. Lee. Algorithmic regularization in learning deep homogeneous models: Layers are automatically balanced. InAdvances in Neural Information Processing Systems, 2018

    2018

  8. [16]

    Fukumizu

    K. Fukumizu. Effect of batch learning in multilayer neural networks. InProceedings of the International Conference on Neural Information Processing, pages 67–70, 1998

    1998

  9. [17]

    Gidel, F

    G. Gidel, F. Bach, and S. Lacoste-Julien. Implicit regularization of discrete gradient dynamics in linear neural networks. InAdvances in Neural Information Processing Systems, 2019

    2019

  10. [18]

    Gissin, S

    D. Gissin, S. Shalev-Shwartz, and A. Daniely. The implicit bias of depth: How incremental learning drives generalization. InProceedings of the International Conference on Learning Representations, 2020

    2020

  11. [19]

    Glorot and Y

    X. Glorot and Y. Bengio. Understanding the difficulty of training deep feedforward neural net- works. InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 249–256. JMLR Workshop and Conference Proceedings, 2010

    2010

  12. [20]

    Golub and C

    G. Golub and C. Van Loan.Matrix Computations. Johns Hopkins Studies in the Mathematical Sciences. Johns Hopkins University Press, 4th edition, 2013

    2013

  13. [21]

    Gunasekar, B

    S. Gunasekar, B. E. Woodworth, S. Bhojanapalli, B. Neyshabur, and N. Srebro. Implicit regular- ization in matrix factorization. InAdvances in Neural Information Processing Systems, volume 30, 2017

    2017

  14. [22]

    Hirsch, R

    M. Hirsch, R. Devaney, and S. Smale.Differential Equations, Dynamical Systems, and Linear Algebra. Pure and Applied Mathematics. Elsevier Science, 1974

    1974

  15. [23]

    Jacot, F

    A. Jacot, F. Ged, B. S ¸im¸ sek, C. Hongler, and F. Gabriel. Saddle-to-saddle dynamics in deep linear networks: Small initialization training, symmetry, and sparsity.arXiv preprint arXiv:2106.15933, 2021

    2021 arXiv

  16. [24]

    Kawaguchi

    K. Kawaguchi. Deep learning without poor local minima. InAdvances in Neural Information Processing Systems, 2016. 44

    2016

  17. [25]

    Khalil.Nonlinear Systems

    H. Khalil.Nonlinear Systems. Pearson Education. Prentice Hall, 2002

    2002

  18. [26]

    Kunin, A

    D. Kunin, A. Ravent´ os, C. Domin´ e, F. Chen, D. Klindt, A. Saxe, and S. Ganguli. Get rich quick: exact solutions reveal how unbalanced initializations promote rapid feature learning. InAdvances in Neural Information Processing Systems, pages 81157–81203, 2024

    2024

  19. [27]

    A. K. Lampinen and S. Ganguli. An analytic theory of generalization dynamics and transfer learning in deep linear networks. InProceedings of the International Conference on Learning Representations, 2019

    2019

  20. [28]

    Laurent and J

    T. Laurent and J. Brecht. Deep linear networks with arbitrary loss: All local minima are global. InProceedings of the International Conference on Machine Learning, pages 2902–2907. PMLR, 2018

    2018

  21. [29]

    LeCun, Y

    Y. LeCun, Y. Bengio, and G. Hinton. Deep learning.Nature, 521(7553):436–444, 2015

    2015

  22. [30]

    J. D. Lee, I. Panageas, G. Piliouras, M. Simchowitz, M. I. Jordan, and B. Recht. First-order methods almost always avoid strict saddle points.Math. Program., 176:311–337, 2019

    2019

  23. [31]

    Z. Li, Y. Luo, and K. Lyu. Towards resolving the implicit bias of gradient descent for matrix fac- torization: Greedy low-rank learning. InProceedings of the International Conference on Learning Representations, 2021

    2021

  24. [32]

    Lojasiewicz

    S. Lojasiewicz. Sur les trajectoires du gradient d’une fonction analytique.Seminari di geometria, 1983:115–117, 1982

    1983

  25. [33]

    H. Min, R. Vidal, and E. Mallada. On the convergence of gradient flow on multi-layer linear models. InProceedings of the International Conference on Machine Learning, pages 24850–24887. PMLR, 2023

    2023

  26. [34]

    Panageas and G

    I. Panageas and G. Piliouras. Gradient descent only converges to minimizers: Non-isolated critical points and invariant regions. In8th Innovations in Theoretical Computer Science Conference (ITCS 2017), pages 2:1–2:12, 2017

    2017

  27. [35]

    Petersen and M

    K. Petersen and M. Petersen. The matrix cookbook.http://matrixcookbook.com, 2012

    2012

  28. [36]

    A. Saxe, J. McClelland, and S. Ganguli. Exact solutions to the nonlinear dynamics of learn- ing in deep linear neural networks. InProceedings of the International Conference on Learning Represenatations 2014, 2014

    2014

  29. [37]

    A. M. Saxe, J. L. McClelland, and S. Ganguli. A mathematical theory of semantic development in deep neural networks.Proc. Natl. Acad. Sci. USA, 116(23):11537–11546, 2019

    2019

  30. [38]

    B.-S. Tam. Circularity of numerical ranges and block-shift matrices.Linear and Multilinear Algebra, 37(1-3):93–109, 1994

    1994

  31. [39]

    Tarmoun, G

    S. Tarmoun, G. Franca, B. D. Haeffele, and R. Vidal. Understanding the dynamics of gradient flow in overparameterized linear models. InProceedings of the International Conference on Machine Learning, pages 10153–10161. PMLR, 2021

    2021

  32. [40]

    Trager, K

    M. Trager, K. Kohn, and J. Bruna. Pure and spurious critical points: a geometric study of linear networks. InProceedings of the International Conference on Learning Representations, 2020

    2020

  33. [41]

    Tsoularis and J

    A. Tsoularis and J. Wallace. Analysis of logistic growth models.Math. Biosci., 179(1):21–55, 2002

    2002

  34. [42]

    Z. Tu, S. T. Aranguri Diaz, and A. Jacot. Mixed dynamics in linear networks: Unifying the lazy and active regimes. InAdvances in Neural Information Processing Systems, pages 106059–106104, 2024

    2024

  35. [43]

    C. Yun, S. Sra, and A. Jadbabaie. Global optimality conditions for deep neural networks. In Proceedings of the International Conference on Learning Representations, 2018. 45

    2018

  36. [44]

    Zhang.Matrix analysis and applications

    X.-D. Zhang.Matrix analysis and applications. Cambridge University Press, 2017

    2017

  37. [45]

    Zhou and Y

    Y. Zhou and Y. Liang. Critical points of neural networks: Analytical forms and landscape prop- erties. InProceedings of the International Conference on Learning Representations, 2018. 46

    2018

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.