Fast and Accurate Homomorphic Softmax Evaluation

• URL: http://arxiv.org/abs/2410.11184v1
• Date: Tue, 15 Oct 2024 02:01:36 GMT
• Title: Fast and Accurate Homomorphic Softmax Evaluation
• Abstract summary: Homomorphic encryption is one of the main solutions for building secure and privacy-preserving solutions for Machine Learning as a Service. We focus on the Softmax function, defined by $mathrmSM(mathbfx) = left(exp(x_i) / sum_j=1n exp(x_j) right)_1le ile n$.
• Score: 2.5766624796160777
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Homomorphic encryption is one of the main solutions for building secure and privacy-preserving solutions for Machine Learning as a Service. This motivates the development of homomorphic algorithms for the main building blocks of AI, typically for the components of the various types of neural networks architectures. Among those components, we focus on the Softmax function, defined by $\mathrm{SM}(\mathbf{x}) = \left(\exp(x_i) / \sum_{j=1}^n \exp(x_j) \right)_{1\le i\le n}$. This function is deemed to be one of the most difficult to evaluate homomorphically, because of its multivariate nature and of the very large range of values for $\exp(x_i)$. The available homomorphic algorithms remain restricted, especially in large dimensions, while important applications such as Large Language Models (LLM) require computing Softmax over large dimensional vectors. In terms of multiplicative depth of the computation (a suitable measure of cost for homomorphic algorithms), our algorithm achieves $O(\log n)$ complexity for a fixed range of inputs, where $n$ is the Softmax dimension. Our algorithm is especially adapted to the situation where we must compute many Softmax at the same time, for instance, in the LLM situation. In that case, assuming that all Softmax calls are packed into $m$ ciphtertexts, the asymptotic amortized multiplicative depth cost per ciphertext is, again over a fixed range, $O(1 + m/N)$ for $N$ the homomorphic ring degree. The main ingredient of our algorithms is a normalize-and-square strategy, which interlaces the exponential computation over a large range and normalization, decomposing both in stabler and cheaper smaller steps. Comparing ourselves to the state of the art, our experiments show, in practice, a good accuracy and a gain of a factor 2.5 to 8 compared to state of the art solutions.

Related papers

• Discretely Beyond $1/e$: Guided Combinatorial Algorithms for Submodular Maximization [13.86054078646307]
  For constrained, not necessarily monotone submodular, all known approximation algorithms with ratio greater than $1/e$ require continuous ideas. For algorithms, the best known approximation ratios for both size and matroid constraint are obtained by a simple randomized greedy algorithm.
  arXiv  Detail & Related papers  (2024-05-08T16:39:59Z)
• Certified Multi-Fidelity Zeroth-Order Optimization [4.450536872346658]
  We consider the problem of multi-fidelity zeroth-order optimization, where one can evaluate a function $f$ at various approximation levels. We propose a certified variant of the MFDOO algorithm and derive a bound on its cost complexity for any Lipschitz function $f$. We also prove an $f$-dependent lower bound showing that this algorithm has a near-optimal cost complexity.
  arXiv  Detail & Related papers  (2023-08-02T07:20:37Z)
• Efficiently Learning One-Hidden-Layer ReLU Networks via Schur Polynomials [50.90125395570797]
  We study the problem of PAC learning a linear combination of $k$ ReLU activations under the standard Gaussian distribution on $mathbbRd$ with respect to the square loss. Our main result is an efficient algorithm for this learning task with sample and computational complexity $(dk/epsilon)O(k)$, whereepsilon>0$ is the target accuracy.
  arXiv  Detail & Related papers  (2023-07-24T14:37:22Z)
• Refined Regret for Adversarial MDPs with Linear Function Approximation [50.00022394876222]
  We consider learning in an adversarial Decision Process (MDP) where the loss functions can change arbitrarily over $K$ episodes. This paper provides two algorithms that improve the regret to $tildemathcal O(K2/3)$ in the same setting.
  arXiv  Detail & Related papers  (2023-01-30T14:37:21Z)
• Private estimation algorithms for stochastic block models and mixture models [63.07482515700984]
  General tools for designing efficient private estimation algorithms. First efficient $(epsilon, delta)$-differentially private algorithm for both weak recovery and exact recovery.
  arXiv  Detail & Related papers  (2023-01-11T09:12:28Z)
• Minimax Optimization with Smooth Algorithmic Adversaries [59.47122537182611]
  We propose a new algorithm for the min-player against smooth algorithms deployed by an adversary. Our algorithm is guaranteed to make monotonic progress having no limit cycles, and to find an appropriate number of gradient ascents.
  arXiv  Detail & Related papers  (2021-06-02T22:03:36Z)
• Empirical Risk Minimization in the Non-interactive Local Model of Differential Privacy [26.69391745812235]
  We study the Empirical Risk Minimization (ERM) problem in the noninteractive Local Differential Privacy (LDP) model. Previous research indicates that the sample complexity, to achieve error $alpha, needs to be depending on dimensionality $p$ for general loss functions.
  arXiv  Detail & Related papers  (2020-11-11T17:48:00Z)
• Hybrid Stochastic-Deterministic Minibatch Proximal Gradient: Less-Than-Single-Pass Optimization with Nearly Optimal Generalization [83.80460802169999]
  We show that HSDMPG can attain an $mathcalObig (1/sttnbig)$ which is at the order of excess error on a learning model. For loss factors, we prove that HSDMPG can attain an $mathcalObig (1/sttnbig)$ which is at the order of excess error on a learning model.
  arXiv  Detail & Related papers  (2020-09-18T02:18:44Z)
• Practical and Parallelizable Algorithms for Non-Monotone Submodular Maximization with Size Constraint [20.104148319012854]
  We present and parallelizable for a submodular function, not necessarily a monotone, with respect to a size constraint. We improve the best approximation factor achieved by an algorithm that has optimal adaptivity and nearly optimal complexity query to $0.193 - varepsilon$.
  arXiv  Detail & Related papers  (2020-09-03T22:43:55Z)
• Revisiting Modified Greedy Algorithm for Monotone Submodular Maximization with a Knapsack Constraint [75.85952446237599]
  We show that a modified greedy algorithm can achieve an approximation factor of $0.305$. We derive a data-dependent upper bound on the optimum. It can also be used to significantly improve the efficiency of such algorithms as branch and bound.
  arXiv  Detail & Related papers  (2020-08-12T15:40:21Z)
• Streaming Complexity of SVMs [110.63976030971106]
  We study the space complexity of solving the bias-regularized SVM problem in the streaming model. We show that for both problems, for dimensions of $frac1lambdaepsilon$, one can obtain streaming algorithms with spacely smaller than $frac1lambdaepsilon$.
  arXiv  Detail & Related papers  (2020-07-07T17:10:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.