betaFeedback

WorksComputer Science

Finite-Time Online Sample Complexity for Conditional Diffusion Models

Abstract

We study the finite-time training and sampling error of conditional diffusion models trained by online stochastic-gradient denoising score matching. Rather than assuming access to an exact empirical or population risk minimizer, we analyze the score network produced after a finite number of stochastic-gradient updates under an explicit localized Polyak–Łojasiewicz (PL) condition, stochastic-gradient regularity, and a trajectory-containment assumption. Our analysis separates the network approximation complexity from the online training-sample budget and propagates both errors through the reverse diffusion process. Under anisotropic smoothness, a joint intrinsic-dimensional structure for the data and conditioning variables, sub-exponential tails, and explicit approximation-class compatibility, we obtain a conditional score approximation error of order O~ ⁣(M−ηapp),ηapp=2αx2αx+dx′+(αx/αy)dy′, \widetilde{\mathcal O}\!\left(M^{-η_{\mathrm{app}}}\right), \qquad η_{\mathrm{app}} = \frac{2α_x} {2α_x+d_x'+(α_x/α_y)d_y'}, where MM denotes the approximation complexity, dx′d_x' and dy′d_y' are the intrinsic response and conditioning dimensions, and αx,αyα_x,α_y are their corresponding smoothness exponents. For N=RBN=RB online denoising samples used by SGD, the finite-time optimization error is controlled explicitly in terms of the PL, smoothness, gradient-noise, and initialization constants. Combining these ingredients with reverse-process stability yields a total-variation bound that decomposes into terminal-initialization, time-discretization, approximation/localization, and finite-time optimization errors.

Introduction

Diffusion and score-based generative models have become a central framework for high-dimensional generative modeling ([29, 16, 30]). Conditional diffusion models extend this framework by learning a family of generative distributions indexed by side information, enabling applications such as text-to-image generation [6, 27], conditional molecular design [24], and conditional time-series imputation [34]. In score-based formulations, the learned object is a conditional score

s⋆(x,t,y)=∇xlog⁡pt(x∣y),s^\star(x,t,y)=\nabla_x\log p_t(x\mid y),

which depends jointly on the noised data variable xx, the conditioning variable yy, and diffusion time tt, and is used to construct the reverse-time generative dynamics. Practical conditional diffusion systems employ a variety of conditioning mechanisms, including classifier guidance [6, 28], classifier-free guidance [15], and transformer-based score architectures [26].

A complete theoretical analysis of conditional diffusion training must connect several distinct components. First, the conditional score must be approximated by the chosen function class. Second, the score-matching objective must be optimized using only a finite amount of computation and data. Third, the resulting score-estimation error must be propagated through a discretized reverse diffusion process. Finally, when the data or conditioning variables are unbounded, the approximation argument must control contributions from regions far from the bulk of the distribution. These issues are coupled in conditional models because approximation takes place jointly over the response, conditioning variable, and diffusion time.

In this work, we study this training-to-sampling problem for conditional diffusion models trained by online stochastic-gradient denoising score matching. Our objective is not to assume that the training procedure returns an exact optimizer and then analyze the resulting score estimator. Instead, we ask how the error of a score network obtained after finitely many stochastic gradient updates propagates to the distribution generated by the reverse process. More precisely, we seek a decomposition in which approximation error, finite-time optimization error, discretization error, and terminal initialization error remain separately visible.

This perspective differs from a common abstraction in statistical analyses of conditional diffusion, where the score estimator is defined through an exact empirical or population risk minimizer [36, 8, 33, 17]. Such an abstraction is useful for isolating score-estimation error, but it does not directly describe the output of a finite stochastic-gradient training procedure. A fixed-dataset analysis may additionally require controlling the generalization gap between empirical and population objectives. Finite-time gradient training has already been analyzed in unconditional settings [13, 14]. Recent work has also analyzed SGD dynamics for unconditional score-based generative models [32]. Thus finite-time gradient training is not, by itself, the novelty claimed here. We instead consider an online sampling model in which each stochastic-gradient update uses fresh draws of the data pair, diffusion time, and Gaussian perturbation. This allows the optimization error to be analyzed directly at the level of the population randomized-time denoising objective.

Replacing exact minimizer access by finite-time optimization requires additional optimization regularity. We make this tradeoff explicit. In particular, our analysis assumes local smoothness, stochastic-gradient regularity, and a localized Polyak–Łojasiewicz (PL) condition along the relevant training trajectory. The PL condition is not derived in this paper for the full neural score class; rather, it is an explicit optimization assumption under which the effect of a finite online SGD budget can be quantified. Thus, the finite-time result should be interpreted as a conditional training guarantee: given the stated localized optimization geometry, we characterize how the remaining optimization error enters the final sampling bound.

The approximation side of the analysis addresses a different difficulty. Conditional score estimation can depend strongly on the dimensions of both the response and the conditioning variable. Some existing guarantees exhibit ambient-dimensional dependence [8], while intrinsic-dimensional analyses exploit manifold or latent low-dimensional structure [33, 17]. We work with a joint intrinsic-dimensional structure for the clean pair (X0,Y)(X_0,Y), together with anisotropic regularity that permits different smoothness along response and conditioning coordinates. We also impose an explicit approximation-class compatibility condition linking our smooth-gated score class to the localized blockwise construction used in the cited conditional approximation theorem; the geometric assumptions alone do not imply the claimed network rate. We denote the corresponding intrinsic dimensions by dx′d'_x and dy′d'_y, and their smoothness exponents by αx\alpha_x and αy\alpha_y, respectively. The associated approximation exponent is

ηapp:=2αx2αx+dx′+(αx/αy)dy′.\eta_{\mathrm{app}} := \frac{2\alpha_x}{2\alpha_x + d'_x + (\alpha_x/\alpha_y)d'_y}.

We additionally allow the data and conditioning variables to have sub-exponential tails. The proof localizes the denoising problem to logarithmic-radius high-probability regions and controls the contribution of their complements through tail decay.

Main result and scope. A central feature of our formulation is that we distinguish the complexity of the score class from the amount of online training data. Let

M=score-network approximation complexity,N=RB=number of online denoising samples,M = \text{score-network approximation complexity}, \qquad N = RB = \text{number of online denoising samples},

where RR is the number of SGD iterations and BB is the mini-batch size. For a discretization grid with KK steps and total time weight HKH_K, our main theorem takes the schematic form

ESGDEY[TV⁡(pt0(⋅∣Y),p^t0(⋅∣Y))]≲Cterme−T+Cdiscrdx+m2K+CstabHK[AK(M)+Eopt(N,M)].\mathbb{E}_{\mathrm{SGD}}\mathbb{E}_Y\left[\operatorname{TV}\left(p_{t_0}(\cdot\mid Y),\widehat{p}_{t_0}(\cdot\mid Y)\right)\right] \lesssim C_{\mathrm{term}}e^{-T} + C_{\mathrm{discr}}\sqrt{\frac{d_x+m_2}{K}} + C_{\mathrm{stab}}\sqrt{H_K\left[\mathcal{A}_K(M)+\mathcal{E}_{\mathrm{opt}}(N,M)\right]}.

under the stated localized PL and trajectory-containment assumptions. Here AK(M)\mathcal{A}_K(M) is the score-class approximation/localization error and Eopt(N,M)\mathcal{E}_{\mathrm{opt}}(N,M) is the excess population denoising risk remaining after finite-time online optimization. Under the anisotropic intrinsic-dimensional approximation assumptions developed in this paper,

AK(M)=O~(M−ηapp).\mathcal{A}_K(M)=\widetilde{O}\left(M^{-\eta_{\mathrm{app}}}\right).

Under localized PL geometry and the decreasing step-size schedule analyzed below, the optimization component satisfies a bound of the form

Eopt(N,M)≲Δ0,Mr02(R+r0)2+LMσM2μM2B(R+r0),N=RB,\mathcal{E}_{\mathrm{opt}}(N,M)\lesssim\frac{\Delta_{0,M}r_0^2}{(R+r_0)^2}+\frac{L_M\sigma_M^2}{\mu_M^2B(R+r_0)}, \qquad N=RB,

where LML_M, μM\mu_M, and σM2\sigma_M^2 denote the relevant smoothness, PL, and stochastic-gradient variance constants, and Δ0,M\Delta_{0,M} is the initial population excess risk. Keeping these quantities explicit separates the approximation and optimization requirements of the result. If these optimization constants remain uniformly controlled along the chosen sequence of score classes, the leading requirements reduce, up to logarithmic factors, to

M=Θ~(ϵ−2/ηapp),N=Θ~(ϵ−2),M=\widetilde{\Theta}\left(\epsilon^{-2/\eta_{\mathrm{app}}}\right), \qquad N=\widetilde{\Theta}\left(\epsilon^{-2}\right),

together with a sufficiently fine reverse-time discretization, for O~(ϵ)\widetilde{O}(\epsilon) expected total-variation error relative to the positive-time target pt0(⋅∣Y)p_{t_0}(\cdot\mid Y). The positive stopping time t0>0t_0>0 is important for interpreting this statement. Our primary result concerns the forward-smoothed conditional law pt0(⋅∣y)p_{t_0}(\cdot\mid y). This permits intrinsic low-dimensional clean-data structure while the reverse-process analysis operates away from the singular endpoint t=0t=0. A total-variation guarantee for the original law p0(⋅∣y)p_0(\cdot\mid y) requires an additional smoothing-bias argument. We provide such an extension under stronger full-dimensional integral density regularity. The stopped-law theorem and the original-law extension therefore have different assumptions and should be viewed as separate results.

ReferenceConditionalIntrinsic dimensionGradient trainingSampling bound
Fu et al. [8]YesNoNoYes
Tang et al. [33]YesYesNoYes
Hu et al. [17]YesYesNoYes
Li [22]YesNoNoYes
Gaur et al. [13]NoNoFinite-time SGDYes
Han et al. [14]NoNoGD with early stoppingNo (training only)
Strasman et al. [32]NoNoFinite-time SGDNo (training only)
This workYesYesOnline SGD under local PLYes

Table 1. Structural comparison with representative adjacent theory. “Gradient training” records whether a concrete finite-run gradient method is analyzed. The rows are not rate comparisons.

Table 1 separates two adjacent lines of work: conditional distribution-estimation theory and algorithm-dependent score-training theory. The assumptions, estimator definitions, and target laws differ, so their rates are not directly comparable. The table locates our contribution at their intersection; it does not assert priority for finite-time diffusion optimization in general.

Contributions. The main contributions of the paper are as follows:

  • Finite-time training-to-sampling decomposition. We give an end-to-end error decomposition for conditional diffusion trained by online stochastic-gradient denoising score matching. The bound keeps terminal initialization, reverse-time discretization, score-class approximation/localization, and finite-time optimization as separate components. In particular, the score network need not be an exact population or empirical minimizer; the remaining optimization error is quantified explicitly under localized PL geometry.

  • Explicit separation of approximation and optimization complexity. We distinguish the score-class complexity MM from the online denoising sample budget NN. Under anisotropic intrinsic-dimensional regularity and the stated approximation-class compatibility condition, the approximation contribution scales as

O~(M−ηapp),ηapp=2αx2αx+dx′+(αx/αy)dy′,\widetilde{O}\left(M^{-\eta_{\mathrm{app}}}\right), \qquad\eta_{\mathrm{app}} = \frac{2\alpha_x}{2\alpha_x+d'_x+(\alpha_x/\alpha_y)d'_y},

while the finite-time optimization term is controlled separately through the smoothness, PL, gradient-noise, initialization, and online-sample parameters. This separation avoids conflating network approximation scale with the number of samples consumed by SGD.

  • Conditional approximation with intrinsic structure and sub-exponential tails. We combine a joint response–conditioning intrinsic geometry with anisotropic smoothness and a sub-exponential tail-localization argument. The resulting construction controls the randomized-time conditional score approximation on logarithmic-radius high-probability regions and then bounds the contribution of the tail complement.

  • Stopped-law guarantee and original-law extension. Our primary theorem controls the learned discretized sampler relative to the forward-smoothed conditional law at a positive stopping time t0t_0. Under additional full-dimensional integral density regularity, we prove a smoothing-bias bound and extend the analysis to total-variation convergence to the original conditional law. This makes explicit where stronger regularity is required and avoids conflating the two guarantees.

At a technical level, the analysis makes compatible three ingredients that are usually treated separately: conditional score approximation, finite-time stochastic optimization, and reverse-diffusion stability. We use existing reverse-process stability and conditional approximation results as inputs where appropriate, while proving the localization, randomized-time approximation transfer, finite-time online-SGD excess-risk control, and their combination into the final training-to-sampling guarantee. Our emphasis is therefore on a transparent and auditable finite-time analysis, including the assumptions under which each component is valid, rather than on claiming a uniformly improved statistical exponent over existing conditional diffusion theory. Relative to unconditional optimization analyses [13, 14, 32], the additional contribution is the conditional coupling: one time-conditioned score must approximate a family indexed by YY, the approximation rate depends on joint fibered geometry, and the resulting randomized-time population error must match the forward-marginal quantity in the conditional reverse-process stability bound. We do not claim a new proof of PL geometry for neural networks or a new reverse-SDE discretization theorem.

Preliminaries and Problem Formulation

Let (X0,Y)∼P0(X_0,Y) \sim P_0 be a random pair, where Y∈RdyY \in\mathbb{R}^{d_y} has marginal law PYP_Y, and, conditional on Y=yY=y, the data variable X0∈RdxX_0 \in\mathbb{R}^{d_x} has conditional law p0(⋅∣y)p_0(\cdot\mid y). Thus

P0(dx,dy)=p0(dx∣y)PY(dy).P_0(dx,dy) = p_0(dx \mid y)P_Y(dy).

We write d=dx+dyd=d_x+d_y. The unconditional setting is recovered by taking dy=0d_y=0. We use pt(⋅∣y)p_t(\cdot\mid y) for a conditional probability law and, when the corresponding density exists, for its density as well.

Throughout, O~\widetilde{O} and Θ~\widetilde{\Theta} suppress only multiplicative constants independent of the displayed asymptotic variables and factors that are polylogarithmic in those variables. Dependence that is polynomial in a vanishing time cutoff such as t0t_0 is displayed explicitly. Fixed distributional, geometric, and moment parameters may enter the implicit constants unless a result states otherwise.

The generative target is the original conditional law p0(⋅∣y)p_0(\cdot\mid y). Our primary theorem, however, controls the learned sampler relative to the forward-smoothed law pt0(⋅∣y)p_{t_0}(\cdot\mid y) at a positive stopping time t0>0t_0>0. Convergence to the original law is treated separately through a smoothing-bias argument under stronger full-dimensional density regularity.

Forward diffusion and conditional score. We consider the variance-preserving Ornstein–Uhlenbeck forward process

dXt=−Xt dt+2 dBt,X0∼p0(⋅∣y),(1)dX_t=-X_t\,dt+\sqrt{2}\,dB_t,\qquad X_0\sim p_0(\cdot\mid y), \tag*{(1)}

where BtB_t is a standard Brownian motion independent of (X0,Y)(X_0,Y). Conditional on X0=x0X_0=x_0, the forward transition can be written as

Xt=e−tX0+σtε,σt2:=1−e−2t,ε∼N(0,Idx),X_t=e^{-t}X_0+\sigma_t\varepsilon,\qquad\sigma_t^2:=1-e^{-2t},\qquad\varepsilon\sim\mathcal{N}(0,I_{d_x}),

with ε\varepsilon independent of (X0,Y)(X_0,Y). Consequently,

qt(x∣x0)=N(e−tx0,σt2Idx),q_t(x\mid x_0)=\mathcal{N}\left(e^{-t}x_0,\sigma_t^2I_{d_x}\right),

and the corresponding kernel score is

∇xlog⁡qt(x∣x0)=−x−e−tx0σt2.(2)\nabla_x\log q_t(x\mid x_0)=-\frac{x-e^{-t}x_0}{\sigma_t^2}. \tag*{(2)}

Let pt(⋅∣y)p_t(\cdot\mid y) denote the conditional law of XtX_t given Y=yY=y. The conditional marginal score is

s∗(x,t,y):=∇xlog⁡pt(x∣y)∈Rdx.(3)s^*(x,t,y):=\nabla_x\log p_t(x\mid y)\in\mathbb{R}^{d_x}. \tag*{(3)}

References

  1. [1]Brian D. O. Anderson. Reverse-time diffusion equation models. Stochastic Processes and their Applications, 12(3):313–326, 1982.DOI
  2. [2]Georgios Batzolis, Jan Stanczuk, Carola-Bibiane Schönlieb, and Christian Etmann. Conditional image generation with score-based diffusion models. arXiv preprint arXiv:2111.13606, 2021.arxiv.org/abs/2111.13606
  3. [3]Joe Benton, Valentin De Bortoli, Arnaud Doucet, and George Deligiannidis. Nearly d-linear convergence bounds for diffusion models via stochastic localization. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=r5njV3Bsud.arxiv.org/abs/2308.03686
  4. [4]Minshuo Chen, Kaixuan Huang, Tuo Zhao, and Mengdi Wang. Score approximation, estimation and distribution recovery of diffusion models on low-dimensional data. In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, pages 4672–4712. PMLR, 2023a.arxiv.org/abs/2302.07194
  5. [6]Prafulla Dhariwal and Alexander Nichol. Diffusion models beat GANs on image synthesis. In Advances in Neural Information Processing Systems, volume 34, pages 8780–8794, 2021.
  6. [8]Hengyu Fu, Zhuoran Yang, Mengdi Wang, and Minshuo Chen. Unveil conditional diffusion models with classifier-free guidance: A sharp statistical theory. arXiv preprint arXiv:2403.11968, 2024.arxiv.org/abs/2403.11968
  7. [9]Mudit Gaur, Vaneet Aggarwal, and Mridul Agarwal. On the global convergence of fitted q-iteration with two-layer neural network parametrization. In Proceedings of the 40th International Conference on Machine Learning, volume 202, pages 11013–11049. PMLR, 23–29 Jul 2023a.arxiv.org/abs/2211.07675
  8. [10]Mudit Gaur, Amrit Singh Bedi, Di Wang, and Vaneet Aggarwal. On the global convergence of natural actor-critic with two-layer neural network parametrization. arXiv preprint arXiv:2306.10486, 2023b.arxiv.org/abs/2306.10486
  9. [11]Mudit Gaur, Amrit Bedi, Di Wang, and Vaneet Aggarwal. Closing the gap: Achieving global convergence (Last iterate) of actor-critic under Markovian sampling with neural network parametrization. In Proceedings of the 41st International Conference on Machine Learning, pages 15153–15179, 2024.arxiv.org/abs/2405.01843
  10. [12]Mudit Gaur, Prashant Trivedi, Shuchin A eron, Amrit Singh Bedi, George K. Atia, and Vaneet Aggarwal. Generative modeling with continuous flows: Sample complexity of flow matching. arXiv preprint arXiv:2512.01286, 2025.arxiv.org/abs/2512.01286
  11. [13]Mudit Gaur, Prashant Trivedi, Sasidhar Kunapuli, Amrit Singh Bedi, and Vaneet Aggarwal. Improved sample complexity bounds for diffusion model training without empirical risk minimizer access. Transactions on Machine Learning Research, 2026. ISSN 2835-8856. URL https://openreview.net/forum?id=CFdNqqlqOv.arxiv.org/abs/2505.18344
  12. [14]Yinbin Han, Meisam Razaviyayn, and Renyuan Xu. Neural network-based score estimation in diffusion models: Optimization and generalization. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=h8GeqOxtd4.arxiv.org/abs/2401.15604
  13. [15]Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022.arxiv.org/abs/2207.12598
  14. [16]Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems, volume 33, pages 6840–6851, 2020.arxiv.org/abs/2006.11239
  15. [17]Jerry Yao-Chieh Hu, Weimin Wu, Yi-Chen Lee, Yu-Chao Huang, Minshuo Chen, and Han Liu. On statistical rates of conditional diffusion transformers: Approximation, estimation and minimax optimality. In The Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=c54apoozCS.arxiv.org/abs/2411.17522
  16. [18]Zhihan Huang, Yuting Wei, and Yuxin Chen. Denoising diffusion probabilistic models are optimally adaptive to unknown low dimensionality. Mathematics of Operations Research, 2026. doi: 10.1287/moor.2024.0769.DOI
  17. [22]Mengze Li. Non-asymptotic convergence bound of conditional diffusion models. arXiv preprint arXiv:2508.10944, 2025.arxiv.org/abs/2508.10944
  18. [23]Jiadong Liang, Zhihan Huang, and Yuxin Chen. Low-dimensional adaptation of diffusion models: Convergence in total variation. arXiv preprint arXiv:2501.12982, 2025.arxiv.org/abs/2501.12982
  19. [24]Aditya Malusare and Vaneet Aggarwal. Improving molecule generation and drug discovery with a knowledge-enhanced generative model. IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2024.DOI
  20. [25]Kazusato Oko, Shunta Akiyama, and Taiji Suzuki. Diffusion models are minimax optimal distribution estimators. In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, pages 26517–26582. PMLR, 2023.arxiv.org/abs/2303.01861
  21. [26]William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205, 2023.arxiv.org/abs/2212.09748
  22. [27]Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10684–10695, 2022.arxiv.org/abs/2112.10752
  23. [28]Rahul Shenoy, Zhihong Pan, Kaushik Balakrishnan, Qisen Cheng, Yongmoon Jeon, Heejune Yang, and Jae-won Kim. Gradient-free classifier guidance for diffusion model sampling. arXiv preprint arXiv:2411.15393, 2024.arxiv.org/abs/2411.15393
  24. [29]Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In Proceedings of the 32nd International Conference on Machine Learning, volume 37 of Proceedings of Machine Learning Research, pages 2256–2265. PMLR, 2015.
  25. [30]Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=PXTiG12RRHS.
  26. [31]Aadithya Srikanth, Mudit Gaur, and Vaneet Aggarwal. Discrete state diffusion models: A sample complexity perspective. In The 29th International Conference on Artificial Intelligence and Statistics, 2026. URL https://openreview.net/forum?id=dTneiZmFre.arxiv.org/abs/2510.10854
  27. [32]Stanislas Strasman, Sobihan Surendran, and Sylvain Le Corff. Non-asymptotic convergence of stochastic gradient descent in score-based generative models. arXiv preprint arXiv:2607.04775, 2026.arxiv.org/abs/2607.04775
  28. [33]Rong Tang, Lizhen Lin, and Yun Yang. Conditional diffusion models are minimax-optimal and manifold-adaptive for conditional distribution estimation. In The Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=NltQraRnbW.arxiv.org/abs/2409.20124
  29. [34]Yusuke Tashiro, Jiaming Song, Yang Song, and Stefano Ermon. CSDI: Conditional score-based diffusion models for probabilistic time series imputation. In Advances in Neural Information Processing Systems, volume 34, pages 24804–24816, 2021.arxiv.org/abs/2107.03502
  30. [35]Andre Wibisono, Yihong Wu, and Kaylee Yingxi Yang. Optimal score estimation via empirical bayes smoothing. In Proceedings of the 37th Conference on Learning Theory, volume 247 of Proceedings of Machine Learning Research, pages 4958–4991. PMLR, 2024.arxiv.org/abs/2402.07747
  31. [36]Hui Yuan, Kaixuan Huang, Chengzhuo Ni, Minshuo Chen, and Mengdi Wang. Reward-directed conditional diffusion: Provable distribution estimation and reward improvement. In Advances in Neural Information Processing Systems, volume 36, 2023. URL https://openreview.net/forum?id=58HwnnEdtF.arxiv.org/abs/2307.07055
  32. [37]Kaihong Zhang, Heqi Yin, Feng Liang, and Jingbo Liu. Minimax optimality of score-based diffusion models: Beyond the density lower bound assumptions. In Proceedings of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine Learning Research. PMLR, 2024.arxiv.org/abs/2402.15602

Paper details

vlnc:W_vychdpye
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents