Pith. sign in

REVIEW 3 major objections 3 minor 39 references

Stochastic Gradient Estimation for Higher-order Differentiable Rendering

T0 review · 3 major / 3 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read By importance sampling a differentiated Gaussian smoothing kernel, this paper derives Monte Carlo estimators for Hessians and Hessian-vector products of the rendering operator, and shows that Newton-conjugate-gradient optimization using…

desk verdict A credible second-order extension of smoothed gradient sampling with a clever O(1) aggregate trick, but sign errors and an unquantified HVP bias need fixing before the numerical claims are trustworthy. read the letter →

arxiv 2412.03489 v3 pith:P5LY45FE submitted 2024-12-04 cs.GR

Zican Wang , Michael Fischer , Tobias Ritschel This is my paper · ORCID
classification cs.GR
keywords differentiablerenderingHessianestimationHessian-vectorproductimportancesamplingsecond-orderoptimizationinverseGaussiansmoothing
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 is trying to establish that second-order derivatives of a rendered image with respect to scene parameters can be estimated without any analytic differentiation of the renderer, purely by Monte Carlo sampling of a smoothed convolution. The trick is to blur the rendering integrand with a Gaussian kernel, differentiate the kernel instead of the scene, and importance-sample the resulting signed distribution; the same recipe yields gradients, Hessians, and Hessian-vector products. Because the renderer is only point-sampled, the method applies to both rasterization and path tracing, and ignores the internal structure of the rendering operator. If correct, this gives inverse rendering practical access to Newton-style optimization, and the paper reports an average 2.71x wall-clock speedup over prior gradient-based methods across its tasks.

What carries the argument

The load-bearing object is the combined smoothing-and-differentiation kernel $D\kappa(\tau)$: a Gaussian convolution kernel with the desired differential operator applied to it, used as the importance-sampling proposal for the convolution integral. Concretely, the gradient kernel is $-\tau_i/\sigma^2 N(\tau,\sigma)$, the Hessian-diagonal kernel is $(-1/\sigma^2+\tau_i^2/\sigma^4)N(\tau,\sigma)$, the Hessian-off-diagonal kernel is $\tau_i\tau_j/\sigma^4 N(\tau,\sigma)$, and the HVP kernel is the central difference of the gradient kernel along the direction $v$. Because the Gaussian factors, the dimensions not being differentiated are sampled independently from a Gaussian. The diagonal-Hessian CDF is transcendental, so its inverse is precomputed in a 1D table for $O(1)$ search; off-diagonals are sampled as products of two independent gradient samples; aggregate sampling substitutes the average over all kernels as a single proposal, justified as a mixture-style combination of unbiased estimators.

What would settle it

On a one-dimensional Gaussian with known second derivative, draw many samples from the tabulated diagonal-Hessian inverse CDF and compare the empirical CDF against the analytic $P^{H}_{ii}(\tau_i)$ in Eq. 21; a persistent mismatch would show the sampler is not faithful to the derived distribution and would break the unbiasedness claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that the differential of the Gaussian-smoothed rendering equation can be written exactly as the product-space integral $$D\bar{L}(x,\omega_o;\$\theta$)=\int_\$\Omega$\int_\Theta D\kappa(\tau)R(\omega_i;\$\theta$-\tau)\,d\tau\,d\omega_i,$$ where $D$ is the differential operator of interest and $\kappa$ is a Gaussian smoothing kernel. Since smoothing makes the integrand differentiable, the derivative passes through the integrals, and the resulting integral is estimated by importance-sampling $D\kappa(\tau)$. For the gradient this reproduces the known smooth-gradient estimator; for the Hessian diagonal the kernel is $(-1/\sigma^2+\tau_i^2/\sigma^4)N(\tau,\sigma)$ and for off-diagonals $\tau_i\tau_j/\sigma^4N(\tau,\sigma)$, each with its own sampling scheme; for Hessian-vector products the estimator is the central difference of the gradient kernel along the direction $v$. The paper further claims that 'aggregate' importance sampling, which draws one sample point from the average of all kernels and reuses it for every matrix element, keeps the estimator unbiased while reducing the cost to a single renderer call. On this basis it reports that Newton-conjugate-gradient optimizers converge faster than gradient descent, by an average factor of 2.71 in wall-clock time.

Load-bearing premise

The load-bearing premise is that the implemented samplers, especially the tabulated inverse CDF for the diagonal Hessian, actually draw from the derived importance distributions, so the Monte Carlo estimates are unbiased in practice and not only in the derivation.

Editorial extensions

If this is right

  • The aggregate Hessian and Hessian-vector product samplers reduce per-iteration renderer calls to a single evaluation, making second-order optimization of rendering problems practical at the same black-box cost as one gradient step.
  • The same black-box point-sampling interface applies to both rasterization and path tracing, so the method does not need renderer-specific differentiation code.
  • Plateau-type inverse rendering problems, where standard differentiable renderers produce zero gradients, become optimizable because the Gaussian smoothing gives the loss landscape a nonzero local slope and curvature.
  • In the paper's evaluations, Newton and conjugate-gradient optimizers using the estimators converge faster in wall-clock time than gradient-descent baselines, with the aggregate HVP variant the fastest overall by an average factor of 2.71.

Reading between the lines

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

  • A natural extension, not tested in the paper, is to apply the same convolution-sampling recipe to other integral operators whose integrands are discontinuous in the parameters, such as differentiable physics simulators.
  • The aggregate sampling trick is a general variance-cost tradeoff: it cuts function evaluations to one but raises variance by sampling from the average kernel; where the crossover makes it unfavorable at high dimension is a testable quantification the paper only begins in its texture task.
  • The grey-box Hessian used in the CNN experiments drops the inner function's second derivative; including that term would test how much of the reported speedup on network tasks comes from true second-order information versus the smoother gradient landscape.
  • Choosing the HVP finite-difference step $\epsilon$ adaptively could reduce the bias of that estimator on problems with heterogeneous curvature.
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

3 major / 3 minor

Summary. The paper proposes a Monte Carlo framework for estimating first- and second-order derivatives of a black-box rendering operator by importance-sampling kernels that combine differentiation with Gaussian smoothing. Starting from the convolution identity in Eq. 15, the authors derive gradient, Hessian, and Hessian-vector-product estimators, introduce an aggregate sampling strategy that reuses one renderer evaluation across all differential dimensions (Table 1), and plug the estimators into Newton-CG and related optimizers. Experiments on analytic, rasterization, path-traced, and CNN-based inverse-rendering tasks report that the aggregate HVP variant converges on average 2.71x faster than previous methods. The supplement contains derivations, an analytic-Hessian validation, and robustness studies with respect to initialization, sample count, MC noise, and JPEG compression.

Significance. If the estimators were unbiased and the implementation faithful, the paper would be a useful practical contribution: it extends differentiable rendering to second-order information at O(1) per-iteration renderer evaluations, and the aggregate sampling idea is a plausible variance/render-cost tradeoff. The core derivation of the Gaussian gradient and Hessian kernels (Secs. 3.2-3.3, Suppl. A-D) is standard and appears correct, and the paper includes a validation against analytic Hessians in Suppl. G and explicit robustness experiments. However, the HVP construction and the inconsistent diagonal-CDF formulas are central to the headline convergence claims, so the significance can only be assessed after those issues are resolved.

major comments (3)
  1. [Sec. 3.4, Eq. (23)] The HVP estimator is not an unbiased estimator of the second-order differential. Eq. (23) defines D_HVP kappa(tau) = [D_G kappa(tau - epsilon v) - D_G kappa(tau + epsilon v)] / (2 epsilon), which is a fixed-epsilon central difference of gradient kernels rather than the exact directional derivative of the Gaussian-gradient kernel. For any fixed epsilon this introduces an O(epsilon^2) bias into the estimated HVP, and the sign is opposite to the central-difference convention used in Eq. (8), since Eq. (8) uses g(theta + epsilon v) - g(theta - epsilon v). As written, the estimator targets -Hv rather than Hv. No bias bound, epsilon schedule, or sensitivity analysis is given, and epsilon does not appear in the hyperparameter table (Table 3), although OursHVP and OursHVPA account for the paper's main convergence speedups. The full-Hessian sampler may be correct, but the HVP-based central claim is unsupported as written.
  2. [Sec. 3.3, Eq. (21) vs Suppl. Sec. D, Eq. (52)] The main text and the supplement give contradictory formulas for the diagonal-Hessian inverse-CDF sampling. Eq. (21) evaluates to 1/4 at u = -sigma, to 3/4 at u = +sigma, and is monotone in the middle interval, which is consistent with the described tail masses. Eq. (52), however, uses P = 1/2 + beta nabla N(tau) on [-sigma, sigma], which gives 3/4 at tau = -sigma and 1/4 at tau = +sigma and is therefore decreasing; the correct middle branch is 1/2 - beta nabla N(tau), and the correct upper branch is 1 + beta nabla N(tau). Since Eq. (21) and Eq. (52) are presented as the same construction and the implementation is described as a tabulation of 'this' CDF, the reader cannot determine which distribution the Hessian sampler actually uses. This directly affects the correctness of OursH and all experiments that use full Hessian information.
  3. [Suppl. Sec. E and Sec. 4.1 (CNN tasks)] The grey-box second-order chain rule is truncated to the Gauss-Newton term: nabla^2_z f(g(x)) is approximated by (nabla_x g)^T nabla^2_y f(g(x)) nabla_x g, omitting the term that contains nabla^2_x g (the second derivative of the inner function). The supplement marks this with 'approximately', but the CNN experiments in Sec. 4.1 use this approximation when reporting higher-order convergence gains. The magnitude of the dropped term is not analyzed for any of the CNN tasks, so the CNN rows in Table 2 do not demonstrate exact second-order information. This is acceptable only if the claim is explicitly limited to a Gauss-Newton approximation; as written, the paper's narrative that higher-order differentials are computed for grey-box compositions is stronger than what Eq. E supports.
minor comments (3)
  1. [Table 2] Numerical footnote markers are embedded inside data cells (e.g., '4', '6', '8', '10' appear within numeric entries), which makes the table difficult to read; these markers should be moved outside the cells or replaced by a legend.
  2. [Table 3] The statement that all hyperparameters, including those of competitors, 'have been optimally chosen' is not backed by a tuning protocol or a sensitivity analysis; given the large variation across tasks, a description of the selection criterion would improve reproducibility.
  3. [Sec. 3.3, paragraph after Eq. (21)] The phrase 'This is a transcendental equation whose inverse CDF cannot be expressed in closed form via elementary functions' is imprecise: the CDF in Eq. (21) is elementary; it is the inverse that requires numerical tabulation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Hessian and HVP estimators are derived from first principles, and the paper includes analytic-Hessian checks.

full rationale

The derivation chain is self-contained. Equation (15) follows from Leibniz' rule after smoothing by convolution (Eqs. 12-14), and the gradient, Hessian, and HVP kernels (Eqs. 17, 20, 23) are computed by explicit differentiation of the Gaussian smoothing kernel in Secs. 3.2-3.4 and Supplemental Secs. A-D. No estimator parameter is fitted to a benchmark output: the smoothing width sigma and HVP step epsilon are user-set hyperparameters (Tab. 3), and the inverse-CDF tables are numerical inversions of analytically derived CDFs (Sec. 3.3), not fits. The paper also validates the sampled Hessian against known analytic Hessians in Supplemental Secs. G and H, providing an external check. The citation of Fischer and Ritschel [7] for first-order gradient sampling is prior work by the same group, but the gradient formula is re-derived in Supplemental Sec. A, and the Hessian/HVP estimators do not inherit their validity from that citation. The HVP estimator is a finite-difference surrogate of the gradient kernel rather than an exact second-derivative kernel (Eq. 23), and the CNN Hessian uses a truncated chain rule (Supplemental Sec. E); these are disclosed approximations that affect bias and correctness, not circular reductions of the claimed result to its own inputs. Therefore no circular step is present.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The method introduces two user-chosen approximation parameters (sigma schedule and HVP epsilon) plus per-task optimizer tuning; the only postulates are standard analytic smoothing assumptions and a truncated chain rule for the CNN experiments.

free parameters (3)
  • Gaussian smoothing width sigma = per-task annealed schedule, e.g., MUG start 3, end 0.01; BUNNY start 1, end 0.01 (Table 3)
    Controls the bias-variance tradeoff of the smoothed derivative estimator; chosen by hand per task and annealed linearly, not derived from first principles. The convergence results depend on the schedule.
  • HVP finite-difference step epsilon = not reported in the provided text
    Eq. 23 uses central differences of the gradient kernel; the step size epsilon is a user choice that controls the bias of the HVP estimate, but its values are not given.
  • Per-task sample count and optimizer hyperparameters = Table 3 lists samples (1-6), trust region, line search iterations, recompute counts per task
    All methods, including competitors, were run with per-task 'optimally chosen' hyperparameters (Table 3 caption), meaning the reported speedups are tuned results, not defaults.
assumptions (5)
  • domain assumption The rendering operator is a black box that can be point-sampled at arbitrary parameter perturbations theta - tau (Sec. 3.1).
    The MC estimator requires evaluating the renderer at displaced scene parameters for each kernel sample; path tracing supports this, rasterizer support is claimed but not detailed.
  • standard math Smoothing by convolution with a Gaussian makes the integrand smooth enough to interchange the differential operator and the integral (Eq. 14, Sec. 3.1).
    Leibniz rule applied to a convolved integrand; valid under integrability conditions taken as given.
  • ad hoc to paper The second-order chain rule for grey-box compositions is truncated to the Gauss-Newton term (supplemental Eq. E).
    For the CNN tasks the Hessian of the composition is approximated as (nabla g)^T H_f (nabla g), dropping the nabla^2 g term; this is acknowledged with approximately-equal but biases the estimated Hessian.
  • domain assumption The inverse CDF of the diagonal Hessian distribution is computed by table lookup and inverse search over a range of 10 sigma (Sec. 3.3).
    The CDF is transcendental (Eq. 21); the tabulation and search are implementation choices that must match the CDF formula. The main text Eq. 21 and supplement Eq. 52 are inconsistent in sign.
  • ad hoc to paper Central differences of the gradient kernel yield a valid HVP kernel (Eq. 23).
    This is a finite-difference approximation of the directional second derivative, not an exact kernel; its O(epsilon^2) bias is not analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stochastic Gradient Estimation for Higher-order Differentiable Rendering." pith.science (2026). https://pith.science/paper/P5LY45FE

@misc{pith2026241203489,
  author       = {Pith},
  title        = {Pith review of: Stochastic Gradient Estimation for Higher-order Differentiable Rendering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P5LY45FE}},
  note         = {Machine review of arXiv:2412.03489}
}
read the original abstract

We derive methods to compute higher order differentials (Hessians and Hessian-vector products) of the rendering operator. Our approach is based on importance sampling of a convolution that represents the differentials of rendering parameters and shows to be applicable to both rasterization and path tracing. We further suggest an aggregate sampling strategy to importance-sample multiple dimensions of one convolution kernel simultaneously. We demonstrate that this information improves convergence when used in higher-order optimizers such as Newton or Conjugate Gradient relative to a gradient descent baseline in several inverse rendering tasks.

Figures

Figures reproduced from arXiv: 2412.03489 by the authors.

Figure 1
Figure 1. Our approach allows sampling the Hessian for inverse [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of classic (yellow) and smooth (red) gradients, as well as our Hessians (blue) on an inverse rendering problem to [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Derivation of smooth differentiation by convolution for [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Aggregate and non-aggregate sampling for a 2D op [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Inverse mesh optimization from renderings. Gradients use [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Detailed plots of the functions involved in the derivation of the CDF for the diagonal elements of [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 8
Figure 8. Figure 8: Optimization on a problem with non-PSD Hessians. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 7
Figure 7. Figure 7: BRDF optimization. the sampled method is less accurate and takes longer due to the sample size and the bandwidth of sampling. The second-order methods outperform the first-order ones in both respects. We show the outcome of our second-order optimization denoted as ’Our…
Figure 9
Figure 9. Figure 9: Overview of all tasks we study (columns): the first row shows the task initialization, the starting point of the optimization. The [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Robustness evaluation (see text). All plots are dis [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 12
Figure 12. Figure 12: Inverse optimization of a heightfield such that the caus [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: Two-bounce mirror optimization experiment. [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    Derivative-free optimization of noisy functions via quasi- newton methods

    Albert S Berahas, Richard H Byrd, and Jorge Nocedal. Derivative-free optimization of noisy functions via quasi- newton methods. SIAM Journal on Optimization, 29(2):965– 993, 2019

    work page 2019

  2. [2]

    Solving transcendental equations: the Cheby- shev polynomial proxy and other numerical rootfinders, per- turbation series, and oracles

    John P Boyd. Solving transcendental equations: the Cheby- shev polynomial proxy and other numerical rootfinders, per- turbation series, and oracles. SIAM, 2014

    work page 2014

  3. [3]

    A nonlinear primal-dual method for total variation-based image restora- tion

    Tony F Chan, Gene H Golub, and Pep Mulet. A nonlinear primal-dual method for total variation-based image restora- tion. SIAM Scient. Comput., 20(6):1964–1977, 1999

    work page 1964

  4. [4]

    Smooth in- terpretation

    Swarat Chaudhuri and Armando Solar-Lezama. Smooth in- terpretation. ACM Sigplan Notices, 45(6):279–291, 2010

    work page 2010

  5. [5]

    Transform- ing a non-differentiable rasterizer into a differentiable one with stochastic gradient estimation

    Thomas Deliot, Eric Heitz, and Laurent Belcour. Transform- ing a non-differentiable rasterizer into a differentiable one with stochastic gradient estimation. In ACM i3D, 2024

    work page 2024

  6. [6]

    Acorns: An easy-to-use code generator for gradients and hessians

    Deshana Desai, Etai Shuchatowitz, Zhongshi Jiang, Teseo Schneider, and Daniele Panozzo. Acorns: An easy-to-use code generator for gradients and hessians. SoftwareX, 17: 100901, 2022

    work page 2022

  7. [7]

    Plateau-reduced differ- entiable path tracing

    Michael Fischer and Tobias Ritschel. Plateau-reduced differ- entiable path tracing. In CVPR, pages 4285–4294, 2023

    work page 2023

  8. [8]

    Zerograds: Learning local surrogates for non-differentiable graphics

    Michael Fischer and Tobias Ritschel. Zerograds: Learning local surrogates for non-differentiable graphics. ACM Trans. Graph. (Proc. SIGGRAPH), 43(4):1–15, 2024

    work page 2024

Show all 39 references
  1. [9]

    A rapidly convergent descent method for minimization

    Roger Fletcher and Michael JD Powell. A rapidly convergent descent method for minimization. The computer journal, 6 (2):163–168, 1963

    1963

  2. [10]

    Adapting arbi- trary normal mutation distributions in evolution strategies: The covariance matrix adaptation

    Nikolaus Hansen and Andreas Ostermeier. Adapting arbi- trary normal mutation distributions in evolution strategies: The covariance matrix adaptation. In Proc. IEEE Evolution- ary Computation, pages 312–317, 1996

    1996

  3. [11]

    Large scale multi-view stereopsis evalu- ation

    Rasmus Jensen, Anders Dahl, George V ogiatzis, Engil Tola, and Henrik Aanæs. Large scale multi-view stereopsis evalu- ation. In CVPR, pages 406–413, 2014

    2014

  4. [12]

    The rendering equation

    James T Kajiya. The rendering equation. In Proc. SIG- GRAPH, 1986

    1986

  5. [13]

    Geo- metric modeling in shape space

    Martin Kilian, Niloy J Mitra, and Helmut Pottmann. Geo- metric modeling in shape space. In SIGGRAPH, pages 64– es, 2007

    2007

  6. [14]

    Fast image deconvolution using hyper-Laplacian priors

    Dilip Krishnan and Rob Fergus. Fast image deconvolution using hyper-Laplacian priors. Proc. NeurIPS, 22, 2009

    2009

  7. [15]

    Soft ras- terizer: A differentiable renderer for image-based 3d reason- ing

    Shichen Liu, Tianye Li, Weikai Chen, and Hao Li. Soft ras- terizer: A differentiable renderer for image-based 3d reason- ing. In Proc. ICCV, 2019

    2019

  8. [16]

    OpenDR: An ap- proximate differentiable renderer

    Matthew M Loper and Michael J Black. OpenDR: An ap- proximate differentiable renderer. In Proc. ECCV, 2014

    2014

  9. [17]

    New insights and perspectives on the natural gradient method

    James Martens. New insights and perspectives on the natural gradient method. J Machine Learning Res , 21(146):1–76, 2020

    2020

  10. [18]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. Comm ACM, 65(1), 2021

    2021

  11. [19]

    Large steps in inverse rendering of geometry

    Baptiste Nicolet, Alec Jacobson, and Wenzel Jakob. Large steps in inverse rendering of geometry. ACM Trans. Graph. (Proc. SIGGRAPH Asis), 40(6):1–13, 2021

    2021

  12. [20]

    Mitsuba 2: A retargetable forward and inverse renderer

    Merlin Nimier-David, Delio Vicini, Tizian Zeltner, and Wen- zel Jakob. Mitsuba 2: A retargetable forward and inverse renderer. ACM Trans. Graph., 38(6), 2019

    2019

  13. [21]

    Radiative backpropagation: an adjoint method for lightning-fast differentiable rendering

    Merlin Nimier-David, S ´ebastien Speierer, Benoˆıt Ruiz, and Wenzel Jakob. Radiative backpropagation: an adjoint method for lightning-fast differentiable rendering. ACM Trans Graph, 39(4), 2020

    2020

  14. [22]

    Numerical optimiza- tion

    Jorge Nocedal and Stephen J Wright. Numerical optimiza- tion. Springer, 1999

    1999

  15. [23]

    Safe and effective importance sam- pling

    Art Owen and Yi Zhou. Safe and effective importance sam- pling. J American Statistical Assoc., 95(449), 2000

    2000

  16. [24]

    Fast exact multiplication by the hes- sian

    Barak A Pearlmutter. Fast exact multiplication by the hes- sian. Neural computation, 6(1):147–160, 1994

    1994

  17. [25]

    Algorithms for nonlinear constraints that use lagrangian functions

    Michael JD Powell. Algorithms for nonlinear constraints that use lagrangian functions. Math. Prog., 14:224–248, 1978

    1978

  18. [26]

    Hierarchical text-conditional image gener- ation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image gener- ation with clip latents. arXiv:2204.06125, 1(2):3, 2022

    2022 arXiv

  19. [27]

    An overview of gradient descent optimiza- tion algorithms

    Sebastian Ruder. An overview of gradient descent optimiza- tion algorithms. arXiv:1609.04747, 2016

    2016 arXiv

  20. [28]

    Texture mapping progressive meshes

    Pedro V Sander, John Snyder, Steven J Gortler, and Hugues Hoppe. Texture mapping progressive meshes. In Proc. SIG- GRAPH, pages 409–416, 2001

    2001

  21. [29]

    Conditioning of quasi-newton methods for function minimization

    David F Shanno. Conditioning of quasi-newton methods for function minimization. Mathematics of computation, 24 (111):647–656, 1970

    1970

  22. [30]

    Adjustment of an inverse matrix correspond- ing to changes in the elements of a given column or row of the original matrix

    Jack Sherman. Adjustment of an inverse matrix correspond- ing to changes in the elements of a given column or row of the original matrix. Annu. Math. Statist., 20:621, 1949

    1949

  23. [31]

    An introduction to the conju- gate gradient method without the agonizing pain

    Jonathan Richard Shewchuk. An introduction to the conju- gate gradient method without the agonizing pain. 1994

    1994

  24. [32]

    Multivariate stochastic approximation using a simultaneous perturbation gradient approximation

    James C Spall. Multivariate stochastic approximation using a simultaneous perturbation gradient approximation. IEEE Trans Automatic Control, 37(3):332–341, 1992

    1992

  25. [33]

    Optimally combining sampling techniques for Monte Carlo rendering

    Eric Veach and Leonidas J Guibas. Optimally combining sampling techniques for Monte Carlo rendering. In Proc. SIGGRAPH, pages 419–428, 1995

    1995

  26. [34]

    Path re- play backpropagation: differentiating light paths using con- stant memory and linear time

    Delio Vicini, S ´ebastien Speierer, and Wenzel Jakob. Path re- play backpropagation: differentiating light paths using con- stant memory and linear time. ACM Trans Graph , 40(4), 2021

    2021

  27. [35]

    Backpropagation: Past and future

    Werbos. Backpropagation: Past and future. In IEEE 1988 Int Conf Neural Networks, pages 343–353. IEEE, 1988

    1988

  28. [36]

    Anisotropic huber-l1 optical flow

    Manuel Werlberger, Werner Trobin, Thomas Pock, Andreas Wedel, Daniel Cremers, and Horst Bischof. Anisotropic huber-l1 optical flow. In BMVC, volume 1, 2009

    2009

  29. [37]

    Adahessian: An adap- tive second order optimizer for machine learning

    Zhewei Yao, Amir Gholami, Sheng Shen, Mustafa Mustafa, Kurt Keutzer, and Michael Mahoney. Adahessian: An adap- tive second order optimizer for machine learning. In Proc. AAAI, volume 35, pages 10665–10673, 2021

    2021

  30. [38]

    Inverse global illumination: Recovering reflectance models of real scenes from photographs

    Yizhou Yu, Paul Debevec, Jitendra Malik, and Tim Hawkins. Inverse global illumination: Recovering reflectance models of real scenes from photographs. InProc. SIGGRAPH, pages 215–224, 1999

    1999

  31. [39]

    A du- ality based approach for realtime TV-L1 optical flow

    Christopher Zach, Thomas Pock, and Horst Bischof. A du- ality based approach for realtime TV-L1 optical flow. In DAGM Pattern Recognition, pages 214–223, 2007. A. Gradient derivation The i-th element of the gradient of the Gaussian, is ∇iN (τ ) = ∂iN (τ ) ∂τi (24) = dN (τi) dτ...

    2007

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.