betaFeedback

WorksComputer Science

Relevance Propagation as Entropic Optimal Transport on Filippov Neural Networks: Sinkhorn Propagation and Region Relevance in Image Classifiers

Abstract

Relevance propagation, which redistributes a network's prediction backward through the layers, is usually treated as a heuristic. We show that it is the solution of a single variational problem: Layer-wise Relevance Propagation (LRP) is the divergence-free flow on the active computational graph of zero relative entropy to the backward transition law of the forward activation flow; for a non-negative reference, this is a relative entropy between Markov measures on backward paths. The rules LRP0_0, LRPz+_{z^+} and LRPα,β_{α,β} follow from this principle by the choice of a reference, and are instances of one rule, LRPNα,β_{α,β}, whose weights may vary from node to node, for instance with the node's activity, in the spirit of Hebb's principle. The principle rests on a measure-theoretic formalization of feed-forward networks as directed acyclic graphs of Filippov operators, which covers piecewise-smooth and discontinuous architectures (e.g. hard-routed Mixture-of-Experts) with a Clarke-Filippov generalized-gradient calculus; it also characterizes backpropagation as the adjoint equation of a Lagrangian relaxation. On networks of ReLU and max-pooling nodes, LRP0_0 is Gradient×\timesInput, with an explicit jump across a kink where LRPz+_{z^+} stays continuous. Taking the entropy edge by edge yields a new rule, Sinkhorn Relevance Propagation. Aggregated over an input region, the flow of a non-negative reference tells, with exact accounting, which hidden units carry its evidence, the conditioned flow being a Schrödinger bridge. On VGG-16, the conservation identities hold to round-off, and the rules are told apart by the sign of their maps, the share of the output left to the bias nodes, faithfulness, localization and class dependence; on ConvNeXt-Tiny, LRPNα,β_{α,β} conserves the output and its Hebbian weights give the highest ABPC of the rules tested.

Introduction

Artificial neural networks routinely outperform humans on narrow, well-specified tasks – from image classification to language modeling – yet the internal mechanisms behind their predictions remain largely opaque. This “black-box” character, present since Rosenblatt [37]’s perceptron, now raises acute concerns of trust, reliability, and accountability wherever these models are deployed in high-stakes decisions – medical diagnosis, autonomous systems, financial decision-making – and has motivated a decade of work in eXplainable Artificial Intelligence (XAI).

Two broad families of attribution methods have emerged to answer the question of why a network produces a given output, each capturing a complementary facet of its behavior. Relevance propagation, epitomized by Layer-wise Relevance Propagation (LRP) (Bach et al., [6]), decomposes the output score as a sum of contributions from individual input features by propagating a conservation signal backward through the network; it tells how the output is shared among the inputs, not how it would respond to a perturbation of them. Gradient-based attribution, such as Gradient×Input [44, 7], quantifies the sensitivity of the output to infinitesimal perturbations; it tells how the output would respond, but ignores the global flow of information through the network. Until now, LRP has been justified by local conservation and by deep Taylor decomposition [30], which derives its rules from local expansions, but not as the solution of a stated optimization problem. In earlier, more applied work [33], we proposed Relative LRP (R-LRP), a rule free of hyperparameters that avoids the divisions by small values of the classical rules by satisfying the conservation law only up to a multiplicative factor, and compared it with the standard LRP variants on image classification with convolutional, VGG and ResNet networks; that experience is what calls for the principled foundation developed here.

This paper replaces the conservation heuristic by a variational principle, which also relates the two families: relevance propagation is exactly the entropic projection of the backward transition law of the forward activation flow onto the set of divergence-free relevance flows on the active computational graph, the unique such flow with zero relative entropy to that law; for a non-negative reference, this relative entropy is the one of two Markov measures on the backward paths of the network. This principle recovers the classical LRPz+\mathrm{LRP}_{z+} and LRPα,β\mathrm{LRP}_{\alpha,\beta} rules as special cases, LRP0\mathrm{LRP}_{0} through a signed extension of the same relative entropy, and gathers the three into one rule whose weights may vary from node to node, for instance with the activity of the node, in the spirit of Hebb’s principle [22]; LRP0\mathrm{LRP}_{0} is Gradient×Input on networks of ReLU and max-pooling nodes. This entropic optimal-transport formulation of relevance is our main contribution: the rules become the choice of a reference, conservation becomes a constraint rather than a heuristic, and the tools of entropic transport – Sinkhorn iterations, Schrödinger bridges – become available for attribution. Carrying this program out rigorously demands a foundation broader than the smooth or piecewise-smooth networks classical LRP was built for: architectures such as hard-routed Mixture-of-Experts [42, 17] can be outright discontinuous, so neither the classical nor the Clarke generalized gradient is everywhere defined, and a gradient-based attribution applied to them naively rests on undefined quantities exactly where the routing decision changes.

The paper is organized as follows. § 2 lays the foundation: a new, measure-theoretic formalization of feed-forward networks as directed acyclic graphs of Filippov operators, with a Clarke–Filippov generalized-gradient calculus yielding well-defined sets of admissible values wherever the network is non-differentiable or discontinuous; recast as a discrete-time optimal control problem in the spirit of LeCun [27], the same formalism gives backpropagation itself a variational characterization, as the adjoint equation of a Lagrangian relaxation. § 3 builds the central variational principle on this foundation, recovering LRPz+\mathrm{LRP}_{z+} and LRPα,β\mathrm{LRP}_{\alpha,\beta} as the unique solutions of relative-entropy minimization problems between backward transition laws, LRP0\mathrm{LRP}_{0} included, its reference being the signed forward contributions themselves. The three are instances of LRPNα,β\mathrm{LRPN}_{\alpha,\beta}, a rule that splits the relevance of each node vv between its excitatory and its inhibitory contributions with weights αv\alpha_v and βv=1−αv\beta_v = 1 - \alpha_v that may change from node to node. With xˉv\bar{x}_v the value of vv divided by the largest value of its layer, LRPNα(xˉ),β(1−xˉ)\mathrm{LRPN}_{\alpha(\bar{x}),\beta(1-\bar{x})} takes αv=xˉv\alpha_v = \bar{x}_v and βv=1−xˉv\beta_v = 1 - \bar{x}_v: in the spirit of Hebb’s principle, a more active node then sends more of its relevance to its excitatory contributions. On networks of ReLU and max-pooling nodes, each of these rules is an input times a modified backward pass – Gradient×Input for LRP0\mathrm{LRP}_{0}, which jumps across a kink where LRPz+\mathrm{LRP}_{z+} stays continuous, whereas every rule may jump where the winner of a max-pooling node changes. Taking instead the relative entropy edge by edge, as in entropically regularized flows, § 3 introduces Sinkhorn Relevance Propagation (SRP), the conservative flow closest to the whole forward activation flow, computed by a Sinkhorn-type iteration, and closes with applications to two image classifiers, VGG-16 and ConvNeXt-Tiny (Appendix § 9). On VGG-16, the rules conserve the output to round-off and are told apart by the sign of their maps, the share of the output they leave to the bias nodes, their faithfulness, their localization and their dependence on the class, SRP following the edges and textures of the lower layers, and a perturbation of standard deviation 10−510^{-5} moves the LRP0\mathrm{LRP}_{0} map by 1.5%, against 0.06 to 0.11% for the other rules, consistently with Corollary 3.5; on ConvNeXt-Tiny, whose hidden values take both signs, LRPNα,β\mathrm{LRPN}_{\alpha,\beta} conserves the output and LRPNα(xˉ),β(1−xˉ)\mathrm{LRPN}_{\alpha(\bar{x}),\beta(1-\bar{x})} has the highest AOPC of the rules tested, whereas LRPz+\mathrm{LRP}_{z+} loses part of the output and LRP0\mathrm{LRP}_{0} is no longer Gradient×Input. § 4 aggregates the relevance flow of a non-negative reference over a region of the input: it tells which hidden units carry the evidence of that region, with exact accounting, the flow conditioned on the region being a Schrödinger bridge whose potentials follow from one forward pass; on 200 annotated photographs of cats and dogs and VGG-16, with LRPz+_{z^+} and LRPNα(xˉ),β(1−xˉ)_{N_{\alpha(\bar{x}),\beta(1-\bar{x})}}, the head receives about 2.5 times as much relevance per pixel as the body, through largely different channels. § 5 concludes and discusses perspectives.

Feed-forward Filippov neural networks

This section lays the foundation of the paper. After the setting and the architectures that motivate it (§ 2.1), it introduces the Filippov regularization of a switching operator (§ 2.2), describes a network as a directed acyclic graph of such operators (§ 2.3), defines its forward propagation, its admissible branch selections, its active graph and the Clarke–Filippov calculus of its output (§ 2.4), and casts backpropagation as the adjoint equation of a discrete-time optimal control problem (§ 2.5).

Setting

The typical setting for supervised learning is to find an approximation of an unknown function ff, given a training data set of its point values: {x(i),f(x(i))}i=1,…,m\{x^{(i)}, f(x^{(i)})\}_{i=1,\ldots,m} where x(i)∈Rdx^{(i)} \in\mathbb{R}^{d} and f(x(i))∈Rpf(x^{(i)}) \in\mathbb{R}^{p}. For example, in classification problems for images, dd is the number of pixels ×\times the number of channels in the images, typically somewhere in the range of 10310^{3} to 10610^{6}, and for videos it is even higher. The learning problem is then to numerically produce from this data a function f^\hat{f} that is in some sense a good predictor of ff on new unseen draws x∈Rdx \in\mathbb{R}^{d}. In the classical formulation the input space is Rd\mathbb{R}^{d}, where each coordinate corresponds to a fixed feature (e.g. a pixel in an image).

Large language models such as Claude [2], Mistral [26, 25], GPT [11, 34], LLaMA [46 b] or DeepSeek [17, 15], are trained on collections of tokenized documents rather than on fixed-dimensional feature vectors. A training sample is a sequence of tokens x=(t1,…,tn)∈Tnx = (t_{1}, \ldots, t_{n}) \in\mathcal{T}^{n}, where T\mathcal{T} is a finite vocabulary and nn may vary from one example to another. The learning task consists in approximating the conditional distribution p(tn+1∣t1,…,tn)p(t_{n+1} \mid t_{1}, \ldots, t_{n}), that is, predicting the next token given the previous context. After embedding, each token is represented by a vector in a latent space Rd\mathbb{R}^{d} and the input to the network becomes a sequence X=(x1,…,xn)∈(Rd)nX = (x_{1}, \ldots, x_{n}) \in(\mathbb{R}^{d})^{n}. Modern foundation models are typically based on transformer architectures combining attention layers, feed-forward networks and normalization operators. Several recent models additionally employ Mixture-of-Experts (MoE) [42] layers in order to increase the number of parameters without increasing the computational cost proportionally. In a MoE block, a routing network computes a score vector. A subset of experts is then selected according to the largest routing scores. In the hard-routing setting, the selection is typically performed through a top-kk operator or an argmax rule.

Unlike attention and feed-forward blocks, which remain smooth or at worst piecewise-smooth (as with ReLU), the hard-routing level selects a different expert map across routing boundaries, where the network map itself, not merely one of its derivatives, may jump, unless the experts agree there. Neither the classical gradient nor the Clarke generalized gradient is then defined at such points, so that a gradient-based explanation applied naively to a hard-routed architecture rests on undefined quantities exactly where the routing decision changes, unless a selection convention is fixed. We therefore introduce, in place of ordinary differentiability, a Filippov set-valued regularization that remains well defined across these discontinuities, coincides with the classical gradient at regular points and contains the Clarke generalized gradient where the network is continuous.

Filippov regularization

Let us introduce piecewise-C1C^{1} switching operator.

Definition 2.1. Let Ω⊆Rn\Omega\subseteq\mathbb{R}^{n} be an open set. A map Φ:Ω∖Σ→Rm\Phi: \Omega\setminus\Sigma\to\mathbb{R}^{m} (m∈N∗m \in\mathbb{N}^{*}) is called a piecewise-C1C^{1} switching operator on Ω\Omega, with switching region Σ\Sigma, if there exist:

  • a locally finite family (Ω(k))k∈K(\Omega^{(k)})_{k \in K} of pairwise disjoint open subsets of Ω\Omega, indexed by a finite or countable set

KK, such that, writing

Σ:=Ω∖⋃k∈KΩ(k),\Sigma:= \Omega\setminus\bigcup_{k\in K}\Omega^{(k)},

the complement satisfies μ(Σ)=0\mu(\Sigma)=0 where μ\mu denotes the Lebesgue measure – an explicit requirement on the family, not automatic from local finiteness and disjointness alone;

  • for every k∈Kk\in K, an open set U(k)⊆RnU^{(k)}\subseteq\mathbb{R}^{n} with Ω(k)∩Ω⊆U(k)\Omega^{(k)}\cap\Omega\subseteq U^{(k)} and a C1C^{1} map Φ(k):U(k)→Rm\Phi^{(k)}:U^{(k)}\to\mathbb{R}^{m}, called the branch of Φ\Phi on Ω(k)\Omega^{(k)}, such that Φ=Φ(k)\Phi=\Phi^{(k)} on Ω(k)\Omega^{(k)}; we write Φ(k)\Phi^{(k)} again for the restriction of this map to Ω(k)∩Ω\Omega^{(k)}\cap\Omega. In particular Φ(k)\Phi^{(k)} is C1C^{1}, with continuous Jacobian DΦ(k)∈Rm×nD\Phi^{(k)}\in\mathbb{R}^{m\times n}, on a full open neighborhood of every point of Ω(k)∩Ω\Omega^{(k)}\cap\Omega, including points of the switching region Σ\Sigma itself, rather than merely admitting a continuous extension from inside Ω(k)\Omega^{(k)}.

For x∈Ωx\in\Omega, the active index set of Φ\Phi at xx is JΦ(x):={k∈K:x∈Ω(k)‾}J_{\Phi}(x):=\{k\in K:x\in\overline{\Omega^{(k)}}\}. We write Φ∈S1(Ω,Rm)\Phi\in S^{1}(\Omega,\mathbb{R}^{m}) for the class of all such operators, and simply Φ∈S1(Ω)\Phi\in S^{1}(\Omega) when m=1m=1.

The set Σ\Sigma is automatically closed in Ω\Omega. Moreover, we will show that JΦ(x)J_{\Phi}(x) is finite and nonempty for every x∈Ωx\in\Omega. Since Φ(k)\Phi^{(k)} is C1C^{1} on the open neighborhood U(k)U^{(k)}, its restriction to Ω(k)‾∩Ω\overline{\Omega^{(k)}}\cap\Omega automatically agrees with the limit taken from inside Ω(k)\Omega^{(k)},

Φ(k)(z)=lim⁡ζ→zζ∈Ω(k)Φ(ζ),for every z∈Ω(k)‾∩Ω,\Phi^{(k)}(z)=\lim_{\substack{\zeta\to z\\\zeta\in\Omega^{(k)}}}\Phi(\zeta),\qquad\text{for every }z\in\overline{\Omega^{(k)}}\cap\Omega,

so the branch is unambiguously determined by Φ\Phi on Ω(k)\Omega^{(k)} alone; the open-neighborhood extension is what additionally guarantees that Φ(k)\Phi^{(k)} stays differentiable, not just continuous, arbitrarily close to a switching point.

References

  1. [1]Reduan Achtibat, Maximilian Dreyer, Ilona Eisenbraun, Sebastian Bosse, Thomas Wiegand, Wojciech Samek, and Sebastian Lapuschkin. From attribution maps to human-understandable explanations through Concept Relevance Propagation. Nature Machine Intelligence, 5(9):1006–1019, 2023.DOI
  2. [2]Anthropic. The Claude 3 model family: Opus, sonnet, haiku. https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf, 2024.
  3. [3]Leila Arras, Ahmed Osman, and Wojciech Samek. CLEVR-XAI: A benchmark dataset for the ground truth evaluation of neural network explanations. Information Fusion, 81:14–40, 2022.DOI
  4. [4]Jean-Pierre Aubin and Arrigo Cellina. Differential Inclusions. Springer, 1984.
  5. [5]Jean-Pierre Aubin and Hélène Frankowska. Set-Valued Analysis. Birkhäuser, 2009.
  6. [6]Sebastian Bach, Alexander Binder, Grégoire Montavon, Frederick Klauschen, Klaus-Robert Müller, and Wojciech Samek. On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PLOS ONE, 10(7):e0130140, 2015. doi: 10.1371/journal.pone.0130140.DOI
  7. [7]David Baehrens, Timon Schroeter, Stefan Harmeling, Motoaki Kawanabe, Katja Hansen, and Klaus-Robert Müller. How to explain individual classification decisions. Journal of Machine Learning Research, 11:1803–1831, 2010. URL https://jmlr.org/papers/v11/baehrens10a.html.arxiv.org/abs/0912.1128
  8. [8]Jean-David Benamou and Yann Brenier. A computational fluid mechanics solution to the monge-kantorovich mass transfer problem. Numer. Math., 84:375–393, 2000. doi: 10.1007/s002110050002.DOI
  9. [9]Mira Bivas, Aris Daniilidis, and Marc Quincampoix. Characterization of filippov representable maps and clarke subdifferentials. Mathematical Programming, 189(1):99–115, 2021. doi: 10.1007/s10107-020-01540-y.DOI
  10. [10]J.A. Bondy and U. S. R. Murty. GRAPH THEORY WITH APPLICATIONS. North-Holland, 1976.
  11. [11]Tom B. Brown et al. Language models are few-shot learners. Advances in Neural Information Processing Systems, 33:1877–1901, 2020.arxiv.org/abs/2005.14165
  12. [12]Lénaïc Chizat, Gabriel Peyré, Bernhard Schmitzer, and François-Xavier Vialard. Scaling algorithms for unbalanced optimal transport problems. Mathematics of Computation, 87(314):2563–2609, 2018. doi: 10.1090/mcom/3303.DOI
  13. [13]Frank H. Clarke. Generalized gradients and applications. Transactions of the American Mathematical Society, 205:247–262, 1975.DOI
  14. [14]Marco Cuturi. Sinkhorn distances: Lightspeed computation of optimal transport. In Advances in Neural Information Processing Systems, volume 26, pages 2292–2300, 2013.
  15. [15]Damai Dai, Chengqi Deng, Chenggang Zhao, R.X. Xu, Huazuo Gao, Deli Chen, Jiashi Li, Wangding Zeng, Xingkai Yu, Y. Wu, et al. Deepseekmoe: Towards ultimate expert specialization in mixture-of-experts language models. arXiv preprint arXiv:2401.06066, 2024.arxiv.org/abs/2401.06066
  16. [16]Ingrid Daubechies, Ron DeVore, Simon Foucart, Boris Hanin, and Guergana Petrova. Nonlinear approximation and (deep) ReLU networks. Constructive Approximation, 55:127–172, 2022. doi: 10.1007/s00365-021-09548-z.DOI
  17. [17]DeepSeek-AI. Deepseek-v3 technical report, 2024. URL https://arxiv.org/abs/2412.19437.
  18. [18]Ronald DeVore, Boris Hanin, and Guergana Petrova. Neural network approximation. Acta Numerica, 30:327–444, 2021. doi: 10.1017/S0962492921000052.DOI
  19. [19]William Fedus, Barret Zoph, and Noam Shazeer. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. Journal of Machine Learning Research, 23(120):1–39, 2022.arxiv.org/abs/2101.03961
  20. [20]A. F. Filippov. Differential Equations with Discontinuous Righthand Sides. Kluwer Academic Publishers, Dordrecht, 1988.
  21. [21]Jindong Gu, Yinchong Yang, and Volker Tresp. Understanding individual decisions of CNNs via contrastive backpropagation. In Asian Conference on Computer Vision (ACCV), pages 119–134, 2018.arxiv.org/abs/1812.02100
  22. [22]Donald Olding Hebb. The organization of behavior: a neuropsychological theory. J. Wiley; Chapman & Hall, 1949.DOI
  23. [23]Jeremy Howard. Imagenette: a smaller subset of 10 easily classified classes from ImageNet. https://github.com/fastai/imagenette, 2019.
  24. [24]Robert A. Jacobs, Michael I. Jordan, Steven J. Nowlan, and Geoffrey E. Hinton. Adaptive mixtures of local experts. Neural Computation, 3(1):79–87, 1991. doi: 10.1162/neco.1991.3.1.79.DOI
  25. [25]Albert Q. Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, et al. Mixtral of experts. arXiv preprint arXiv:2401.04088, 2024.arxiv.org/abs/2401.04088
  26. [26]Albert Q. Jiang et al. Mistral 7b. arXiv preprint arXiv:2310.06825, 2023.arxiv.org/abs/2310.06825
  27. [27]Yann LeCun. A theoretical framework for back-propagation. In David Touretzky, Geoffrey Hinton, and Terrence Sejnowski, editors, Proceedings of the 1988 Connectionist Models Summer School, pages 21–28, Pittsburgh, PA, 1988.
  28. [28]Matthias Liero, Alexander Mielke, and Giuseppe Savaré. Optimal entropy-transport problems and a new hɛllinger-kantorovich distance between positive measures. Inventiones mathematicae, 211(3):969–1117, 2018. doi: 10.1007/s00222-017-0759-8.
  29. [29]Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. A ConvNet for the 2020s. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11976–11986, 2022.
  30. [30]Grégoire Montavon, Sebastian Lapuschkin, Wojciech Samek, Alexander Binder, and Klaus-Robert Müller. Explaining nonlinear classification decisions with deep taylor decomposition. Pattern Recogn., 65:211–222, 2017.
  31. [31]Grégoire Montavon, Wojciech Samek, and Klaus-Robert Müller. Methods for interpreting and understanding deep neural networks. Digital Signal Processing, 73:1–15, 2018. doi: 10.1016/j.dsp.2017.10.011.
  32. [32]Grégoire Montavon, Alexander Binder, Sebastian Lapuschkin, Wojciech Samek, and Klaus-Robert Müller. Layer-Wise Relevance Propagation: An Overview, volume 11700, pages 193–209. Springer, 2019. ISBN 978-3-030-28953-9. doi: 10.1007/978-3-030-28954-6_10.DOI
  33. [33]Eric Nyiri and Olivier Gibaru. Relative layer-wise relevance propagation: a more robust neural networks explanation, 2025. arXiv:2501.14322.arxiv.org/abs/2501.14322
  34. [34]OpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023.arxiv.org/abs/2303.08774
  35. [35]Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. Cats and dogs. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 3498–3505, 2012.
  36. [36]Gabriel Peyré and Marco Cuturi. Computational optimal transport: With applications to data science. Foundations and Trends in Machine Learning, 11(5–6):355–607, 2019. doi: 10.1561/2200000073.
  37. [37]F. Rosenblatt. The perceptron: a probabilistic model for information storage and organization in the brain. Psychological review, 65(6):386–408, 1958. doi: 10.1037/h0042519.
  38. [38]Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet large scale visual recognition challenge. International Journal of Computer Vision, 115(3):211–252, 2015.
  39. [39]Wojciech Samek, Alexander Binder, Grégoire Montavon, Sebastian Lapuschkin, and Klaus-Robert Müller. Evaluating the visualization of what a deep neural network has learned. IEEE Transactions on Neural Networks and Learning Systems, 28(11):2660–2673, 2017.DOI
  40. [40]Stefan Scholtes. Introduction to Piecewise Differentiable Equations. SpringerBriefs in Optimization. Springer, New York, 2012.DOI
  41. [41]Thibault Séjourné, Gabriel Peyré, and François-Xavier Vialard. Unbalanced optimal transport, from theory to numerics. In Handbook of Numerical Analysis, volume 24, pages 407–471. Elsevier, 2023.arxiv.org/abs/2211.08775
  42. [42]Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. In International Conference on Learning Representations (ICLR), 2017.arxiv.org/abs/1701.06538
  43. [43]Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. In International Conference on Learning Representations (ICLR), 2015.
  44. [44]Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps. In Workshop at the International Conference on Learning Representations (ICLR), 2014. URL https://arxiv.org/abs/1312.6034.
  45. [45]Leon Sixt, Maximilian Granz, and Tim Landgraf. When explanations lie: Why many modified BP attributions fail. In International Conference on Machine Learning (ICML), pages 9046–9057, 2020.arxiv.org/abs/1912.09818
  46. [46]Hugo Touvron et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023a.arxiv.org/abs/2302.13971
  47. [48]Cédric Villani. Optimal Transport: Old and New, volume 338 of Grundlehren der mathematischen Wissenschaften. Springer, 2009. doi: 10.1007/978-3-540-71050-9.

Paper details

vlnc:W_mhvvcnkn
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