Pith. sign in

REVIEW 4 major objections 6 minor 66 references

L-SR1: Learned Symmetric-Rank-One Preconditioning

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read L-SR1, a compact learned second-order optimizer, reports PA-MPJPE 51.58 on 3DPW by generating SR1-style preconditioners that are trained to satisfy the quasi-Newton secant relation.

desk verdict A solid empirical learned-optimizer paper that overstates its 'learned projection'—the actual method is a secant-regularized SR1-style preconditioner, and the no-annotated-data claim is contradicted by Eq. 13. read the letter →

arxiv 2508.12270 v3 pith:DQI3Y77U submitted 2025-08-17 cs.LG cs.CV

Gal Lifshitz , Shahar Zuler , Ori Fouks , Dan Raviv This is my paper · ORCID
classification cs.LGcs.CV MSC 90C5365K0568T07
keywords learnedoptimizationsecond-orderquasi-Newtonmethodssymmetricrank-onepreconditioninghumanmeshrecoverymeta-learning
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

L-SR1 is a learned second-order optimizer that extends the classical Symmetric Rank-One method: a small neural network produces vectors whose outer products precondition gradient updates, and a training-time penalty pushes those preconditioners toward the quasi-Newton secant relation. The paper's central claim is that this compact, self-supervised design can converge faster and to better solutions than both classical optimizers and prior learned optimizers on analytic benchmarks and on monocular human mesh recovery. On the 3DPW benchmark the paper reports PA-MPJPE 51.58 for L-SR1, versus 55.90 for the LGD baseline it replaces, with a smaller model and no task-specific fine-tuning. If true, this would make learned second-order preconditioning a practical drop-in component for optimization-based fitting pipelines, including in settings where dimension changes across problems.

What carries the argument

The load-bearing object is the preconditioner $\tilde{B}_k = B_0 + \sum_{i=1}^L v_i v_i^{\top}$, where $B_0 = I$ and each $v_i$ is generated by a shared multilayer perceptron called the Vector Generator from an encoded optimization state and kept in a fixed-size buffer. The outer-product form automatically makes $\tilde{B}_k$ symmetric positive semi-definite, which turns every update into a descent direction. The secant condition $B_k q_k = p_k$, the quasi-Newton requirement that the approximate inverse Hessian map the change in gradient to the change in parameters, is converted into a projection objective and then into the training penalty $\lambda_{\text{sec}}\lVert p_k - \tilde{B}_k q_k \rVert_2^2$ in the meta-loss. That penalty is the mechanism meant to make the learned preconditioner behave like an inverse-Hessian approximation at inference, even though no explicit projection step is run during optimization.

What would settle it

Run a trained L-SR1 checkpoint on held-out quadratics and record, for each inner iteration, the normalized secant residual $\lVert p_k - \tilde{B}_k q_k \rVert_2 / \lVert p_k \rVert_2$. If that residual is no smaller than for a model trained with $\lambda_{\text{sec}} = 0$, the projection mechanism is not doing the claimed work and the performance gains come from the learned preconditioning alone.

Watch

Extended reading notes

Core claim

The paper's central claim is that a learned preconditioner built from rank-one outer products can transfer the strengths of quasi-Newton methods into a trainable optimizer without sacrificing stability. At each iteration L-SR1 stores the last $L$ vectors produced by a learned vector generator and computes the descent direction as $d_k = \sum_{v \in \mathcal{B}_L} v v^{\top} g_{k-1}$; because the buffer stores only vectors, the effective preconditioner $\tilde{B}_k = I + \sum_i v_i v_i^{\top}$ is always symmetric positive semi-definite, guaranteeing a descent direction. The quasi-Newton structure is imposed by a projection objective, minimizing $\lVert p_k - \tilde{B}_k q_k \rVert_2^2$, which is implemented as a secant penalty in the meta-loss rather than as an exact projection at inference. The reported consequences are that L-SR1 obtains the highest performance profile among six solvers on a 30-problem analytic suite and, on 3DPW, lowers PA-MPJPE from 55.90 (LGD) to 51.58 with a 10.4M-parameter model that needs no annotated 3D data or fine-tuning.

Load-bearing premise

The load-bearing premise is that the training-time penalty that pushes the learned preconditioner toward the quasi-Newton secant relation, the requirement that the approximate inverse Hessian map the change in gradient to the change in parameters, is strong enough to make that relation hold on test problems, where no exact projection is applied.

Editorial extensions

If this is right

  • A single L-SR1 checkpoint, trained on quadratics and two benchmark functions in dimension 100, attains the top performance profile on a test suite with dimensions 50 to 1000, indicating the elementwise design transfers across problem sizes.
  • Replacing the LGD update module with L-SR1 in a human mesh recovery pipeline improves PA-MPJPE on 3DPW from 55.90 to 51.58, with a smaller model of 10.4M versus 17.4M parameters, so learned second-order preconditioning is competitive in a high-dimensional vision task.
  • Because all learnable modules operate elementwise and the vector buffer has a fixed size, memory use grows only linearly with dimension and per-iteration runtime is reported as lower than LGD's (91 ms versus 166 ms in the computational appendix).
  • The self-supervised training on a 2D reprojection loss plus the secant penalty means the optimizer does not need annotated 3D meshes or per-task fine-tuning, lowering the barrier to inserting it into other iterative fitting pipelines.

Reading between the lines

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

  • Editorial inference: if the secant penalty is what carries the method, then sweeping $\lambda_{\text{sec}}$ on the HMR task should reproduce the accuracy gap that the paper shows with and without the projection on quadratics, but the paper does not report that sweep.
  • Editorial inference: the same preconditioner could be tested as a drop-in update rule for non-convex problems such as neural network training, where the paper's analytic evidence does not directly apply.
  • Editorial inference: the method's dimension invariance suggests one trained optimizer could be shared across a family of tasks of very different sizes, but the paper evaluates transfer from a single training dimension rather than multi-task training on mixed dimensions.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes L-SR1, a learned second-order optimizer that uses MLP-generated vectors to build a limited-memory preconditioner of the form B_k = I + sum_i v_i v_i^T, together with elementwise learned learning rates. The headline component, the "Projection-Guided Secant Mechanism" (PGSM), is described as a learned projection that enforces the secant condition and positive semidefiniteness. The method is evaluated on random quadratics, on performance profiles over benchmark functions, and on monocular Human Mesh Recovery (HMR), where Table 1 reports PA-MPJPE 51.58 for L-SR1 versus 55.90 for LGD on 3DPW. The paper claims state-of-the-art results among optimization-based HMR methods and strong generalization across problem dimensions without fine-tuning.

Significance. If fully supported, the paper would contribute a compact, dimension-invariant learned second-order preconditioner that can be dropped into optimization pipelines such as HMR. The empirical setup is broad, including analytic benchmarks, performance profiles, runtime and memory measurements, ablations, and a real-world integration. The PSD construction via outer products is simple and cheap, and the runtime comparison in Appendix A is useful. However, the two headline properties are not actually implemented as stated: the secant condition is only softly penalized in the meta-loss and is not enforced at inference, and the claimed self-supervision is contradicted by the ground-truth term in the HMR meta-loss. The HMR comparison also rests on a single unpaired number. As it stands, the evidence supports a weaker claim: a learned PSD preconditioner regularized toward the secant relation, which may still be useful but is not the "learned projection" advertised.

major comments (4)
  1. [Sec. 4.2, Eq. (8)-(9); Algorithm 1] The central "learned projection" is never computed at inference. Eq. (8) defines a projection as an argmin over the structured family, but Eq. (9) replaces it by adding the soft penalty lambda_sec * ||p_k - B_k q_k||^2 to the meta-loss. Algorithm 1 (lines 4-11) uses the raw vector-generator output v_k = P(f_k) directly in the buffer and computes d_k = sum_{v in B_L} v v^T g_{k-1}; no projected vector is applied. Thus the claim that L-SR1 "enforces" the secant condition is not implemented. This is load-bearing because the method's name, the second contribution bullet, and the quasi-Newton grounding in Sec. 3.2 all rest on this mechanism. The authors should either apply an actual projection at inference or revise the claims and report the test-time secant residual ||p_k - B_k q_k|| to show that the soft penalty suffices.
  2. [Sec. 5.2, Eq. (13); Abstract; Contributions] The paper repeatedly describes the method as self-supervised and claims it works "without the need for annotated data or supervised meta-training," but Eq. (13) contains the term lambda_self * ||Theta_k - Theta_gt||_1, where Theta_gt are ground-truth SMPL parameters from AMASS. This is supervised meta-training with annotated ground truth. The claim should be corrected, or the loss and training protocol must be changed so that no ground-truth parameters are used.
  3. [Table 1; Sec. 5.2] The HMR comparison is reported as a single number per method with no standard deviations, no number of seeds, and no significance testing. L-SR1's headline value of 51.58 is reached after 13 steps, while LGD's 55.90 is reached after 4; the bracketed value 51.74 after 4 steps is the relevant point comparison, but even that is one run. The statement that L-SR1 "consistently outperforms" LGD in 3D accuracy is therefore not statistically supported. Multiple seeds and an error bar or significance test are needed.
  4. [Sec. 5.1.2; Fig. 3c; Table 6] The performance-profile claim that L-SR1 achieves the highest profile is presented without repeated runs, confidence intervals, or sensitivity analysis. The profile depends on the tuned learning rates of the non-trainable baselines and on the buffer sizes and secant weights chosen in Table 6, so the claimed consistent effectiveness needs robustness evidence across these choices.
minor comments (6)
  1. [Table 1] The header contains a typo: "PA-MPJEPE" should be "PA-MPJPE".
  2. [Appendix C.2.1; Appendix C.2.3] The symbol gamma is used for both the learning-rate generator exponent scale and the meta-learning-rate decay; rename one of them to avoid ambiguity.
  3. [References] Reference [29] is missing venue and year information and should be completed.
  4. [Eq. (13)] The symbol R_sec is used in Eq. (13) without being defined in that equation; please restate or point explicitly to Eq. (9) at first use.
  5. [Fig. 3 caption] The caption says "with and without learned projection," but neither variant applies a projection at inference; the difference is the presence of the secant penalty in the meta-loss. Please rephrase to avoid implying that the projection is executed.
  6. [Reproducibility] The paper states that code will be released upon acceptance, but currently provides no seeds or exact train/validation checkpoint selection details; these should be included to make the single-number HMR result reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: L-SR1's central claims are empirical and evaluated against external benchmarks, and its training objective, while including the secant penalty, is not used to derive the reported performance.

full rationale

The paper's load-bearing claims—that L-SR1 achieves PA-MPJPE 51.58 on 3DPW and the highest performance profile on analytic benchmarks—are measured outcomes, not quantities obtained by construction from the method's inputs. The secant penalty R_sec in Eq. (9) is a soft training regularizer; it is not claimed to be a test-time projection, and the reported results are not derived from the magnitude of that penalty. The PSD property holds by construction (B0 = I plus a sum of outer products), but this is a structural guarantee, not a prediction derived from fitted data. The paper does not fit any parameter to the target benchmarks and then rename that fit as a prediction; all HMR and analytic results are against held-out or external test sets. The manuscript contains no load-bearing self-citations, no imported uniqueness theorems, and no ansatz smuggled via citation. The gap between the claimed 'learned projection' and the implemented penalty is a correctness/verification concern, not a circularity: even if the secant residual is never checked at inference, the empirical improvement of the projected variant over the non-projected variant is a genuine experimental finding. Therefore, the derivation chain is self-contained and no circular step can be exhibited from the paper's own equations or citations.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The central claim rests on several hand-chosen hyperparameters (gamma_1, gamma_2, lambda_sec, buffer size, hidden dimension) and on the unproven assumption that a soft secant penalty transfers to hard secant satisfaction at inference. The only invented entity, the learned projection, is described as a projection but is implemented only as a loss term.

free parameters (5)
  • gamma_1 (learning-rate generator scale) = 0.4 for quadratic, 0.1 for profiles and HMR
    Hand-chosen scaling for the elementwise learning rate; appears in Sec. 4.1 and Appendix C.
  • gamma_2 (learning-rate generator exponent scale) = 0.001 for all experiments
    Hand-chosen; controls the spread of coordinate-wise learning rates.
  • lambda_sec (secant penalty weight) = 100 (quadratic), 10 (profile quadratics), 1 (Rosenbrock, Rastrigin, HMR)
    Hand-chosen weight for the secant-condition penalty in the meta-loss (Eq. 9, Table 6).
  • Buffer size L = 8 (quadratic), 16/32 (profile training), 64 (profile test), 4 (HMR)
    Hand-chosen number of retained rank-one vectors; affects the preconditioner's capacity.
  • Hidden dimension d_hidden = 128
    Hand-chosen width of the MLP modules (Appendix C.1); ablations tested 64 and 128.
assumptions (4)
  • standard math The objective function is twice continuously differentiable (f in C^2), so the secant equation follows from a first-order Taylor expansion of the gradient.
    Invoked in Sec. 3.2 and Lemma 1; standard background for quasi-Newton methods.
  • standard math B0 = I and B_tilde_k = B0 + sum_i v_i v_i^T is positive semidefinite.
    Sum of a positive definite matrix and PSD rank-one terms; used in Sec. 4.2.
  • ad hoc to paper Minimizing the secant penalty R_sec during meta-training yields secant satisfaction at inference.
    The paper claims the learned projection satisfies the secant condition, but at inference the network only generates vectors; the penalty is soft and there is no enforcement step (Sec. 4.2, Eqs. 8-9).
  • ad hoc to paper Elementwise MLPs preserve dimension invariance and are expressive enough to learn useful cross-coordinate curvature information through outer products.
    The architecture (Sec. 4.1) processes each coordinate independently; the paper provides no analysis of how cross-coordinate structure is learned.
invented entities (1)
  • Learned projection operation (Eq. 6, Eq. 8)
    purpose: Purported to map any rank-one-updated preconditioner to the nearest PSD matrix satisfying the secant condition.
    No such projection is computed in the implementation; the method only adds a training penalty R_sec (Eq. 9), so the entity as described does not exist at inference.

how reviews work

0 comments
Cite this review

Pith. "Pith review of L-SR1: Learned Symmetric-Rank-One Preconditioning." pith.science (2026). https://pith.science/paper/DQI3Y77U

@misc{pith2026250812270,
  author       = {Pith},
  title        = {Pith review of: L-SR1: Learned Symmetric-Rank-One Preconditioning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DQI3Y77U}},
  note         = {Machine review of arXiv:2508.12270}
}
read the original abstract

End-to-end deep learning has achieved impressive results but often relies on large labeled datasets, exhibits limited generalization to unseen scenarios, and incurs substantial computational cost. Classical optimization methods, in contrast, are more data-efficient and lightweight but frequently suffer from slow convergence. Learned optimizers aim to bridge this gap, yet existing approaches have focused primarily on first-order methods, while learned second-order optimization has received much less attention. We introduce L-SR1, a learned second-order optimizer inspired by the classical Symmetric Rank-One (SR1) method. At its core, L-SR1 employs a Projection-Guided Secant Mechanism (PGSM) that generates positive semi-definite preconditioners and biases meta-training toward the quasi-Newton secant relation. Through controlled analytic benchmarks, we study stability, generalization across problem dimensions, and search direction quality, and further evaluate L-SR1 on Monocular Human Mesh Recovery (HMR), where it outperforms both classical and learned optimization-based baselines. With a compact model and no reliance on task-specific fine-tuning or annotated data, L-SR1 demonstrates strong generalization and can be integrated into a broad range of iterative optimization problems to accelerate convergence and reduce the required number of iterations.

Figures

Figures reproduced from arXiv: 2508.12270 by the authors.

Figure 1
Figure 1. Optimization trajectories. Our evaluation spans both classic analytic functions and the real-world human mesh recovery (HMR). Shown here are example optimization trajectories on a quadratic function and two well-known challenging benchmark functions [56]—the Rosenbrock and Rastrigin functions. In this example, we compare LBFGS with our lightweight L-SR1 method, with and without the proposed learned projection. The l… view at source ↗
Figure 2
Figure 2. Learned-SR1 (L-SR1) iteration block diagram. At each iteration k, the Input Encoder E receives the vectors xk−1, pk−1, dk−1, gk−1, and qk−1, producing a feature vector fk. This is passed to the Vector Generator P, which outputs a new direction vector vk, and to the Learning Rate Generator G, which produces element-wise learning rates αk (Sec. 4.1)). The updated descent direction dk is computed as a sum of rank-one t… view at source ↗
Figure 3
Figure 3. Analytic optimization experiments. Figures 3a and 3b show results from our quadratic experiments (Section 5.1.1), comparing L-SR1 (with and without learned projection) to LBFGS [43], which serves as a reference. Figures 3a and 3b report objective values and cosine similarities with the Newton direction, respectively, averaged over the test set. The learned projection improves L-SR1’s convergence and its alignment wi… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: HMR error curves on 3DPW. We compare 15 inner iterations of LGD [54], L-SR1 (ours), and LBFGS [31], which is initialized using our trained initialization module. Fig. 4a shows 2D joint reprojection error; Fig. 4b shows PA-MPJPE. Dots mark the suggested stopping steps. …
Figure 5
Figure 5. Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Computational effort during inference. Measurements were collected on a single NVIDIA RTX 3090 GPU and correspond to the mean runtime per inner inference iteration and peak memory with a batch size of 32 [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Computational effort during meta-training. Measurements were collected on a single NVIDIA RTX 3090 GPU and correspond to the mean runtime per inner inference iteration and peak memory usage with a batch size of 4 [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: HMR meta-training on AMASS and validation erros on 3DPW. Shown first 100K iterations. Our meta-loss is higher as it has the secant loss added to it. iteration, a batch of 2D joints is generated by projecting the corresponding 3D joints onto a randomly sampled camera vi…
Figure 9
Figure 9. Figure 9: Qualitative comparison of L-SR1 (ours) and LGD HMR results. Meshes optimized with L-SR1 (Ours) are shown in white, while those from LGD [54] are shown in pink. Regions of interest are highlighted with red circles. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 48 canonical work pages

  1. [17]

    Transformer- based learned optimization

    Erik Gärtner, Luke Metz, Mykhaylo Andriluka, C Daniel Freeman, and Cristian Sminchisescu. Transformer- based learned optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11970–11979, 2023

    work page 2023

  2. [1]

    ACCAD MoCap Dataset

    Advanced Computing Center for the Arts and Design. ACCAD MoCap Dataset

  3. [2]

    Ijaz Akhter and Michael J. Black. Pose-conditioned joint angle limits for 3D human pose reconstruction. In IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) 2015, June 2015

    work page 2015

  4. [3]

    Learning to learn by gradient descent by gradient descent

    Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew W Hoffman, David Pfau, Tom Schaul, Brendan Shillingford, and Nando De Freitas. Learning to learn by gradient descent by gradient descent. Advances in neural information processing systems, 29, 2016

    2016

  5. [4]

    Digital dance ethnography: Organizing large dance collections

    Andreas Aristidou, Ariel Shamir, and Yiorgos Chrysanthou. Digital dance ethnography: Organizing large dance collections. J. Comput. Cult. Herit., 12(4), Nov. 2019

    work page 2019

  6. [5]

    Qn-mixer: A quasi-newton mlp-mixer model for sparse- view ct reconstruction

    Ishak Ayad, Nicolas Larue, and Maï K Nguyen. Qn-mixer: A quasi-newton mlp-mixer model for sparse- view ct reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 25317–25326, 2024

    work page 2024

  7. [6]

    Best practices for comparing optimization algorithms

    Vahid Beiranvand, Warren Hare, and Yves Lucet. Best practices for comparing optimization algorithms. Optimization and Engineering, 18:815–848, 2017

    work page 2017

  8. [7]

    Bertsekas

    D.P. Bertsekas. Nonlinear Programming. Athena Scientific, 1999

    work page 1999

Show all 66 references
  1. [8]

    Keep it smpl: Automatic estimation of 3d human pose and shape from a single image

    Federica Bogo, Angjoo Kanazawa, Christoph Lassner, Peter Gehler, Javier Romero, and Michael J Black. Keep it smpl: Automatic estimation of 3d human pose and shape from a single image. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 1...

    2016

  2. [9]

    Federica Bogo, Javier Romero, Gerard Pons-Moll, and Michael J. Black. Dynamic FAUST: Registering human bodies in motion. In IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), July 2017

    2017

  3. [10]

    Convex optimization

    Stephen Boyd, Stephen P Boyd, and Lieven Vandenberghe. Convex optimization. Cambridge university press, 2004

    2004

  4. [11]

    Kemp, and James Hays

    Samarth Brahmbhatt, Cusuh Ham, Charles C. Kemp, and James Hays. ContactDB: Analyzing and predicting grasp contact via thermal imaging. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019

    2019

  5. [12]

    Openpose: Realtime multi- person 2d pose estimation using part affinity fields

    Zhe Cao, Gines Hidalgo, Tomas Simon, Shih-En Wei, and Yaser Sheikh. Openpose: Realtime multi- person 2d pose estimation using part affinity fields. IEEE transactions on pattern analysis and machine intelligence, 43(1):172–186, 2019

    2019

  6. [13]

    CMU MoCap Dataset

    Carnegie Mellon University. CMU MoCap Dataset

  7. [14]

    Human4d: A human-centric multimodal dataset for motions and immersive media

    Anargyros Chatzitofis, Leonidas Saroglou, Prodromos Boutis, Petros Drakoulis, Nikolaos Zioulis, Shishir Subramanyam, Bart Kevelham, Caecilia Charbonnier, Pablo Cesar, Dimitrios Zarpalas, et al. Human4d: A human-centric multimodal dataset for motions and immersive media. IEEE A...

    2020

  8. [15]

    Convergence of quasi-newton matrices generated by the symmetric rank one update

    Andrew R Conn, Nicholas IM Gould, and Ph L Toint. Convergence of quasi-newton matrices generated by the symmetric rank one update. Mathematical programming, 50(1):177–195, 1991

    1991

  9. [16]

    Benchmarking optimization software with performance profiles

    Elizabeth D Dolan and Jorge J Moré. Benchmarking optimization software with performance profiles. Mathematical programming, 91:201–213, 2002

    2002

  10. [18]

    Nima Ghorbani and Michael J. Black. SOMA: Solving optical marker-based mocap automatically. In Proc. International Conference on Computer Vision (ICCV), pages 11117–11126, Oct. 2021

    2021

  11. [19]

    Saeed Ghorbani, Kimia Mahdaviani, Anne Thaler, Konrad Kording, Douglas James Cook, Gunnar Blohm, and Nikolaus F. Troje. MoVi: A large multipurpose motion and video dataset. arXiv preprint arXiv: 2003.01888, 2020

    2003 arXiv

  12. [20]

    Shampoo: Preconditioned stochastic tensor optimization

    Vineet Gupta, Tomer Koren, and Yoram Singer. Shampoo: Preconditioned stochastic tensor optimization. In International Conference on Machine Learning, pages 1842–1850. PMLR, 2018

    2018

  13. [21]

    Bewegungsanalyse getäuschter und nicht-getäuschter 7m-würfe im handball

    Fabian Helm, Nikolaus Troje, Mathias Reiser, and Jörn Munzert. Bewegungsanalyse getäuschter und nicht-getäuschter 7m-würfe im handball. 01 2015

    2015

  14. [22]

    Sleight of hand: Perception of finger motion from reduced marker sets

    Ludovic Hoyet, Kenneth Ryall, Rachel McDonnell, and Carol O’Sullivan. Sleight of hand: Perception of finger motion from reduced marker sets. In Proceedings of the ACM SIGGRAPH Symposium on Interactive 3D Graphics and Games, pages 79–86, 2012

    2012

  15. [23]

    Mnemosyne: Learning to train transformers with transformers

    Deepali Jain, Krzysztof M Choromanski, Kumar Avinava Dubey, Sumeet Singh, Vikas Sindhwani, Tingnan Zhang, and Jie Tan. Mnemosyne: Learning to train transformers with transformers. Advances in Neural Information Processing Systems, 36:77331–77358, 2023

    2023

  16. [24]

    A theoretical and experimental study of the symmetric rank-one update

    H Fayez Khalfan, Richard H Byrd, and Robert B Schnabel. A theoretical and experimental study of the symmetric rank-one update. SIAM Journal on Optimization, 3(1):1–24, 1993

    1993

  17. [25]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  18. [26]

    Learning to reconstruct 3d human pose and shape via model-fitting in the loop

    Nikos Kolotouros, Georgios Pavlakos, Michael J Black, and Kostas Daniilidis. Learning to reconstruct 3d human pose and shape via model-fitting in the loop. In Proceedings of the IEEE/CVF international conference on computer vision, pages 2252–2261, 2019. 10

    2019

  19. [27]

    The KIT bimanual manipulation dataset

    Franziska Krebs, Andre Meixner, Isabel Patzer, and Tamim Asfour. The KIT bimanual manipulation dataset. In IEEE/RAS International Conference on Humanoid Robots (Humanoids), pages 499–506, 2021

    2021

  20. [28]

    Learning to optimize

    Ke Li and Jitendra Malik. Learning to optimize. arXiv preprint arXiv:1606.01885, 2016

    2016 arXiv

  21. [29]

    Learning to combine quasi-newton methods

    Maojia Li, Jialin Liu, and Wotao Yin. Learning to combine quasi-newton methods

  22. [30]

    Wheelposer: Sparse-imu based body pose estimation for wheelchair users

    Yunzhi Li, Vimal Mollyn, Kuang Yuan, and Patrick Carrington. Wheelposer: Sparse-imu based body pose estimation for wheelchair users. In Proceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility, pages 1–17, 2024

    2024

  23. [31]

    On the limited memory method for large scale optimization: Mathematical programming b

    DC Liu and J Nocedal. On the limited memory method for large scale optimization: Mathematical programming b. 1989

    1989

  24. [32]

    Smpl: A skinned multi-person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi-person linear model. In Seminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 851–866. 2023

    2023

  25. [33]

    Loper, Naureen Mahmood, and Michael J

    Matthew M. Loper, Naureen Mahmood, and Michael J. Black. MoSh: Motion and shape capture from sparse markers. ACM Transactions on Graphics, (Proc. SIGGRAPH Asia) , 33(6):220:1–220:13, Nov. 2014

    2014

  26. [34]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  27. [35]

    Eyes JAPAN Co. Ltd. Eyes Japan MoCap Dataset

  28. [36]

    Troje, Gerard Pons-Moll, and Michael J

    Naureen Mahmood, Nima Ghorbani, Nikolaus F. Troje, Gerard Pons-Moll, and Michael J. Black. AMASS: Archive of motion capture as surface shapes. In International Conference on Computer Vision, pages 5442–5451, Oct. 2019

    2019

  29. [37]

    The KIT whole-body human motion database

    Christian Mandery, Ömer Terlemez, Martin Do, Nikolaus Vahrenkamp, and Tamim Asfour. The KIT whole-body human motion database. In International Conference on Advanced Robotics (ICAR), pages 329–336, 2015

    2015

  30. [38]

    Unifying rep- resentations and large-scale whole-body motion databases for studying human motion

    Christian Mandery, Ömer Terlemez, Martin Do, Nikolaus Vahrenkamp, and Tamim Asfour. Unifying rep- resentations and large-scale whole-body motion databases for studying human motion. IEEE Transactions on Robotics, 32(4):796–809, 2016

    2016

  31. [39]

    Optimizing neural networks with kronecker-factored approximate curvature

    James Martens and Roger Grosse. Optimizing neural networks with kronecker-factored approximate curvature. In International conference on machine learning, pages 2408–2417. PMLR, 2015

    2015

  32. [40]

    Tasks, stability, architecture, and compute: Training more effective learned optimizers, and using them to train themselves

    Luke Metz, Niru Maheswaranathan, C Daniel Freeman, Ben Poole, and Jascha Sohl-Dickstein. Tasks, stability, architecture, and compute: Training more effective learned optimizers, and using them to train themselves. arXiv preprint arXiv:2009.11243, 2020

    2009 arXiv

  33. [41]

    Müller, T

    M. Müller, T. Röder, M. Clausen, B. Eberhardt, B. Krüger, and A. Weber. Documentation mocap database hdm05. Technical Report CG-2007-2, Universität Bonn, June 2007

    2007

  34. [42]

    A method of solving a convex programming problem with convergence rate o\bigl(kˆ2\bigr)

    Yurii Evgen’evich Nesterov. A method of solving a convex programming problem with convergence rate o\bigl(kˆ2\bigr). In Doklady Akademii Nauk, volume 269, pages 543–547. Russian Academy of Sciences, 1983

    1983

  35. [43]

    Updating quasi-newton matrices with limited storage

    Jorge Nocedal. Updating quasi-newton matrices with limited storage. Mathematics of computation , 35(151):773–782, 1980

    1980

  36. [44]

    Ahmed A A Osman, Timo Bolkart, and Michael J. Black. STAR: A sparse trained articulated human body regressor. In European Conference on Computer Vision (ECCV), pages 598–613, 2020

    2020

  37. [45]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  38. [46]

    Expressive body capture: 3d hands, face, and body from a single image

    Georgios Pavlakos, Vasileios Choutas, Nima Ghorbani, Timo Bolkart, Ahmed AA Osman, Dimitrios Tzionas, and Michael J Black. Expressive body capture: 3d hands, face, and body from a single image. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognitio...

    2019

  39. [47]

    Kalmannet: Neural network aided kalman filtering for partially known dynamics

    Guy Revach, Nir Shlezinger, Xiaoyong Ni, Adria Lopez Escoriza, Ruud JG Van Sloun, and Yonina C Eldar. Kalmannet: Neural network aided kalman filtering for partially known dynamics. IEEE Transactions on Signal Processing, 70:1532–1547, 2022

    2022

  40. [48]

    A deterministic global optimization using smooth diagonal auxiliary functions

    Yaroslav D Sergeyev and Dmitri E Kvasov. A deterministic global optimization using smooth diagonal auxiliary functions. Communications in Nonlinear Science and Numerical Simulation, 21(1-3):99–111, 2015

    2015

  41. [49]

    Pliks: A pseudo-linear inverse kinematic solver for 3d human body estimation

    Karthik Shetty, Annette Birkhold, Srikrishna Jaganathan, Norbert Strobel, Markus Kowarschik, Andreas Maier, and Bernhard Egger. Pliks: A pseudo-linear inverse kinematic solver for 3d human body estimation. In Proceedings of the IEEE/CVF conference on computer vision and patter...

    2023

  42. [50]

    Wham: Reconstructing world-grounded humans with accurate 3d motion

    Soyong Shin, Juyong Kim, Eni Halilaj, and Michael J Black. Wham: Reconstructing world-grounded humans with accurate 3d motion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2070–2080, 2024

    2024

  43. [51]

    Viterbinet: A deep learning based viterbi algorithm for symbol detection

    Nir Shlezinger, Nariman Farsad, Yonina C Eldar, and Andrea J Goldsmith. Viterbinet: A deep learning based viterbi algorithm for symbol detection. IEEE Transactions on Wireless Communications, 19(5):3319– 3331, 2020. 11

    2020

  44. [52]

    Model-based deep learning

    Nir Shlezinger, Jay Whang, Yonina C Eldar, and Alexandros G Dimakis. Model-based deep learning. Proceedings of the IEEE, 2023

    2023

  45. [53]

    Sigal, A

    L. Sigal, A. Balan, and M. J. Black. HumanEva: Synchronized video and motion capture dataset and baseline algorithm for evaluation of articulated human motion. International Journal of Computer Vision, 87(1):4–27, Mar. 2010

    2010

  46. [54]

    Human body model fitting by learned gradient descent

    Jie Song, Xu Chen, and Otmar Hilliges. Human body model fitting by learned gradient descent. In European Conference on Computer Vision, pages 744–760. Springer, 2020

    2020

  47. [55]

    Trace: 5d temporal regression of avatars with dynamic cameras in 3d environments

    Yu Sun, Qian Bao, Wu Liu, Tao Mei, and Michael J Black. Trace: 5d temporal regression of avatars with dynamic cameras in 3d environments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8856–8866, 2023

    2023

  48. [56]

    Surjanovic and D

    S. Surjanovic and D. Bingham. Virtual library of simulation experiments: Test functions and datasets. Retrieved May 5, 2024, from http://www.sfu.ca/~ssurjano

    2024

  49. [57]

    On the importance of initialization and momentum in deep learning

    Ilya Sutskever, James Martens, George Dahl, and Geoffrey Hinton. On the importance of initialization and momentum in deep learning. In International conference on machine learning, pages 1139–1147. PMLR, 2013

    2013

  50. [58]

    Black, and Dimitrios Tzionas

    Omid Taheri, Nima Ghorbani, Michael J. Black, and Dimitrios Tzionas. GRAB: A dataset of whole-body human grasping of objects. In European Conference on Computer Vision (ECCV), 2020

    2020

  51. [59]

    Huang, Taheri Omid, Michael J

    Shashank Tripathi, Lea Müller, Chun-Hao P. Huang, Taheri Omid, Michael J. Black, and Dimitrios Tzionas. 3D human pose estimation via intuitive physics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023

    2023

  52. [60]

    Nikolaus F. Troje. Decomposing biological motion: A framework for analysis and synthesis of human gait patterns. Journal of Vision, 2(5):2–2, Sept. 2002

    2002

  53. [61]

    Total Capture: 3d human pose estimation fusing video and inertial sensors

    Matt Trumble, Andrew Gilbert, Charles Malleson, Adrian Hilton, and John Collomosse. Total Capture: 3d human pose estimation fusing video and inertial sensors. In 2017 British Machine Vision Conference (BMVC), 2017

    2017

  54. [62]

    SFU Motion Capture Database

    Simon Fraser University and National University of Singapore. SFU Motion Capture Database

  55. [63]

    Recover- ing accurate 3d human pose in the wild using imus and a moving camera

    Timo von Marcard, Roberto Henschel, Michael Black, Bodo Rosenhahn, and Gerard Pons-Moll. Recover- ing accurate 3d human pose in the wild using imus and a moving camera. In European Conference on Computer Vision (ECCV), sep 2018

    2018

  56. [64]

    Prompthmr: Promptable human mesh recovery

    Yufu Wang, Yu Sun, Priyanka Patel, Kostas Daniilidis, Michael J Black, and Muhammed Kocabas. Prompthmr: Promptable human mesh recovery. arXiv preprint arXiv:2504.06397, 2025

    2025 arXiv

  57. [65]

    Learned optimizers that scale and generalize

    Olga Wichrowska, Niru Maheswaranathan, Matthew W Hoffman, Sergio Gomez Colmenarejo, Misha Denil, Nando Freitas, and Jascha Sohl-Dickstein. Learned optimizers that scale and generalize. In International conference on machine learning, pages 3751–3760. PMLR, 2017

    2017

  58. [66]

    secant constraint

    Zhewei Yao, Amir Gholami, Sheng Shen, Mustafa Mustafa, Kurt Keutzer, and Michael Mahoney. Adahes- sian: An adaptive second order optimizer for machine learning. In proceedings of the AAAI conference on artificial intelligence, volume 35, pages 10665–10673. 12 A Computational A...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.