亚洲男人的天堂2018av,欧美草比,久久久久久免费视频精选,国色天香在线看免费,久久久久亚洲av成人片仓井空

Hyperparameter tuning is a common practice in the application of machine learning but is a typically ignored aspect in the literature on privacy-preserving machine learning due to its negative effect on the overall privacy parameter. In this paper, we aim to tackle this fundamental yet challenging problem by providing an effective hyperparameter tuning framework with differential privacy. The proposed method allows us to adopt a broader hyperparameter search space and even to perform a grid search over the whole space, since its privacy loss parameter is independent of the number of hyperparameter candidates. Interestingly, it instead correlates with the utility gained from hyperparameter searching, revealing an explicit and mandatory trade-off between privacy and utility. Theoretically, we show that its additional privacy loss bound incurred by hyperparameter tuning is upper-bounded by the squared root of the gained utility. However, we note that the additional privacy loss bound would empirically scale like a squared root of the logarithm of the utility term, benefiting from the design of doubling step.

The formal privacy guarantee provided by Differential Privacy (DP) bounds the leakage of sensitive information from deep learning models. In practice, however, this comes at a severe computation and accuracy cost. The recently established state of the art (SOTA) results in image classification under DP are due to the use of heavy data augmentation and large batch sizes, leading to a drastically increased computation overhead. In this work, we propose to use more efficient models with improved feature quality by introducing steerable equivariant convolutional networks for DP training. We demonstrate that our models are able to outperform the current SOTA performance on CIFAR-10 by up to $9\%$ across different $\varepsilon$-values while reducing the number of model parameters by a factor of $35$ and decreasing the computation time by more than $90 \%$. Our results are a large step towards efficient model architectures that make optimal use of their parameters and bridge the privacy-utility gap between private and non-private deep learning for computer vision.

We introduce the concurrent shuffle model of differential privacy. In this model we have multiple concurrent shufflers permuting messages from different, possibly overlapping, batches of users. Similarly to the standard (single) shuffle model, the privacy requirement is that the concatenation of all shuffled messages should be differentially private. We study the private continual summation problem (a.k.a. the counter problem) and show that the concurrent shuffle model allows for significantly improved error compared to a standard (single) shuffle model. Specifically, we give a summation algorithm with error $\tilde{O}(n^{1/(2k+1)})$ with $k$ concurrent shufflers on a sequence of length $n$. Furthermore, we prove that this bound is tight for any $k$, even if the algorithm can choose the sizes of the batches adaptively. For $k=\log n$ shufflers, the resulting error is polylogarithmic, much better than $\tilde{\Theta}(n^{1/3})$ which we show is the smallest possible with a single shuffler. We use our online summation algorithm to get algorithms with improved regret bounds for the contextual linear bandit problem. In particular we get optimal $\tilde{O}(\sqrt{n})$ regret with $k= \tilde{\Omega}(\log n)$ concurrent shufflers.

Traditionally, the random noise is equally injected when training with different data instances in the field of differential privacy (DP). In this paper, we first give sharper excess risk bounds of DP stochastic gradient descent (SGD) method. Considering most of the previous methods are under convex conditions, we use Polyak-{\L}ojasiewicz condition to relax it in this paper. Then, after observing that different training data instances affect the machine learning model to different extent, we consider the heterogeneity of training data and attempt to improve the performance of DP-SGD from a new perspective. Specifically, by introducing the influence function (IF), we quantitatively measure the contributions of various training data on the final machine learning model. If the contribution made by a single data instance is so little that attackers cannot infer anything from the model, we do not add noise when training with it. Based on this observation, we design a `Performance Improving' DP-SGD algorithm: PIDP-SGD. Theoretical and experimental results show that our proposed PIDP-SGD improves the performance significantly.

Consensus and leader election are fundamental problems in distributed systems. Consensus is the problem in which all processes in a distributed computation must agree on some value. Average consensus is a popular form of consensus, where the agreed upon value is the average of the initial values of all the processes. In a typical solution for consensus, each process learns the value of others' to determine the final decision. However, this is undesirable if processes want to keep their values secret from others. With this motivation, we present a solution to privacy-preserving average consensus, where no process can learn the initial value of any other process. Additionally, we augment our approach to provide outlier resistance, where extreme values are not included in the average calculation. Privacy is fully preserved at every stage, including preventing any process from learning the identities of processes that hold outlier values. To our knowledge, this is the first privacy-preserving average consensus algorithm featuring outlier resistance. In the context of leader election, each process votes for the one that it wants to be the leader. The goal is to ensure that the leader is elected in such a way that each vote remains secret and the sum of votes remain secret during the election. Only the final vote tally is available to all processes. This ensures that processes that vote early are not able to influence the votes of other processes. We augment our approach with shallow ranked voting by allowing processes to not only vote for a single process, but to designate a secondary process to vote towards in the event that their primary vote's candidate does not win the election.

In this paper, we present a unified and general framework for analyzing the batch updating approach to nonlinear, high-dimensional optimization. The framework encompasses all the currently used batch updating approaches, and is applicable to nonconvex as well as convex functions. Moreover, the framework permits the use of noise-corrupted gradients, as well as first-order approximations to the gradient (sometimes referred to as "gradient-free" approaches). By viewing the analysis of the iterations as a problem in the convergence of stochastic processes, we are able to establish a very general theorem, which includes most known convergence results for zeroth-order and first-order methods. The analysis of "second-order" or momentum-based methods is not a part of this paper, and will be studied elsewhere. However, numerical experiments indicate that momentum-based methods can fail if the true gradient is replaced by its first-order approximation. This requires further theoretical analysis.

Consider a pair of random variables $(X,Y)$ distributed according to a given joint distribution $p_{XY}$. A curator wishes to maximally disclose information about $Y$, while limiting the information leakage incurred on $X$. Adopting mutual information to measure both utility and privacy of this information disclosure, the problem is to maximize $I(Y;U)$, subject to $I(X;U)\leq\epsilon$, where $U$ denotes the released random variable and $\epsilon$ is a given privacy threshold. Two settings are considered, where in the first one, the curator has access to $(X,Y)$, and hence, the optimization is over $p_{U|XY}$, while in the second one, the curator can only observe $Y$ and the optimization is over $p_{U|Y}$. In both settings, the utility-privacy trade-off is investigated from theoretical and practical perspective. More specifically, several privacy-preserving schemes are proposed in these settings based on generalizing the notion of statistical independence. Moreover, closed-form solutions are provided in certain scenarios. Finally, convexity arguments are provided for the utility-privacy trade-off as functionals of the joint distribution $p_{XY}$.

Federated Learning is a distributed machine learning environment, which ensures that clients complete collaborative training without sharing private data, only by exchanging parameters. However, the data does not satisfy the same distribution and the computing resources of clients are different, which brings challenges to the related research. To better solve the above heterogeneous problems, we designed a novel federated learning method. The local model consists of the pre-trained model as the backbone and fully connected layers as the head. The backbone can extract features for the head, and the embedding vector of classes is shared between clients to optimize the head so that the local model can perform better. By sharing the embedding vector of classes, instead of parameters based on gradient space, clients can better adapt to private data, and it is more efficient in the communication between the server and clients. To better protect privacy, we proposed a privacy-preserving hybrid method to add noise to the embedding vector of classes, which has less impact on the local model performance under the premise of satisfying differential privacy. We conduct a comprehensive evaluation with other federated learning methods on the self-built vehicle dataset under non-independent identically distributed(Non-IID)

In this paper, we study a distributed privacy-preserving learning problem in social networks with general topology. The agents can communicate with each other over the network, which may result in privacy disclosure, since the trustworthiness of the agents cannot be guaranteed. Given a set of options which yield unknown stochastic rewards, each agent is required to learn the best one, aiming at maximizing the resulting expected average cumulative reward. To serve the above goal, we propose a four-staged distributed algorithm which efficiently exploits the collaboration among the agents while preserving the local privacy for each of them. In particular, our algorithm proceeds iteratively, and in every round, each agent i) randomly perturbs its adoption for the privacy-preserving purpose, ii) disseminates the perturbed adoption over the social network in a nearly uniform manner through random walking, iii) selects an option by referring to the perturbed suggestions received from its peers, and iv) decides whether or not to adopt the selected option as preference according to its latest reward feedback. Through solid theoretical analysis, we quantify the trade-off among the number of agents (or communication overhead), privacy preserving and learning utility. We also perform extensive simulations to verify the efficacy of our proposed social learning algorithm.

Classic algorithms and machine learning systems like neural networks are both abundant in everyday life. While classic computer science algorithms are suitable for precise execution of exactly defined tasks such as finding the shortest path in a large graph, neural networks allow learning from data to predict the most likely answer in more complex tasks such as image classification, which cannot be reduced to an exact algorithm. To get the best of both worlds, this thesis explores combining both concepts leading to more robust, better performing, more interpretable, more computationally efficient, and more data efficient architectures. The thesis formalizes the idea of algorithmic supervision, which allows a neural network to learn from or in conjunction with an algorithm. When integrating an algorithm into a neural architecture, it is important that the algorithm is differentiable such that the architecture can be trained end-to-end and gradients can be propagated back through the algorithm in a meaningful way. To make algorithms differentiable, this thesis proposes a general method for continuously relaxing algorithms by perturbing variables and approximating the expectation value in closed form, i.e., without sampling. In addition, this thesis proposes differentiable algorithms, such as differentiable sorting networks, differentiable renderers, and differentiable logic gate networks. Finally, this thesis presents alternative training strategies for learning with algorithms.