When training Convolutional Neural Networks (CNNs) there is a large emphasis on creating efficient optimization algorithms and highly accurate networks. The state-of-the-art method of optimizing the networks is done by using gradient descent algorithms, such as Stochastic Gradient Descent (SGD). However, there are some limitations presented when using gradient descent methods. The major drawback is the lack of exploration, and over-reliance on exploitation. Hence, this research aims to analyze an alternative approach to optimizing neural network (NN) weights, with the use of population-based metaheuristic algorithms. A hybrid between Grey Wolf Optimizer (GWO) and Genetic Algorithms (GA) is explored, in conjunction with SGD; producing a Genetically Modified Wolf optimization algorithm boosted with SGD (GMW-SGD). This algorithm allows for a combination between exploitation and exploration, whilst also tackling the issue of high-dimensionality, affecting the performance of standard metaheuristic algorithms. The proposed algorithm was trained and tested on CIFAR-10 where it performs comparably to the SGD algorithm, reaching high test accuracy, and significantly outperforms standard metaheuristic algorithms.
相關內容
Nonlinear control systems with partial information to the decision maker are prevalent in a variety of applications. As a step toward studying such nonlinear systems, this work explores reinforcement learning methods for finding the optimal policy in the nearly linear-quadratic regulator systems. In particular, we consider a dynamic system that combines linear and nonlinear components, and is governed by a policy with the same structure. Assuming that the nonlinear component comprises kernels with small Lipschitz coefficients, we characterize the optimization landscape of the cost function. Although the cost function is nonconvex in general, we establish the local strong convexity and smoothness in the vicinity of the global optimizer. Additionally, we propose an initialization mechanism to leverage these properties. Building on the developments, we design a policy gradient algorithm that is guaranteed to converge to the globally optimal policy with a linear rate.
This paper explores continuous-time control synthesis for target-driven navigation to satisfy complex high-level tasks expressed as linear temporal logic (LTL). We propose a model-free framework using deep reinforcement learning (DRL) where the underlying dynamic system is unknown (an opaque box). Unlike prior work, this paper considers scenarios where the given LTL specification might be infeasible and therefore cannot be accomplished globally. Instead of modifying the given LTL formula, we provide a general DRL-based approach to satisfy it with minimal violation. To do this, we transform a previously multi-objective DRL problem, which requires simultaneous automata satisfaction and minimum violation cost, into a single objective. By guiding the DRL agent with a sampling-based path planning algorithm for the potentially infeasible LTL task, the proposed approach mitigates the myopic tendencies of DRL, which are often an issue when learning general LTL tasks that can have long or infinite horizons. This is achieved by decomposing an infeasible LTL formula into several reach-avoid sub-tasks with shorter horizons, which can be trained in a modular DRL architecture. Furthermore, we overcome the challenge of the exploration process for DRL in complex and cluttered environments by using path planners to design rewards that are dense in the configuration space. The benefits of the presented approach are demonstrated through testing on various complex nonlinear systems and compared with state-of-the-art baselines. The Video demonstration can be found here://youtu.be/jBhx6Nv224E.
The conventional machine learning (ML) and deep learning approaches need to share customers' sensitive information with an external credit bureau to generate a prediction model that opens the door to privacy leakage. This leakage risk makes financial companies face an enormous challenge in their cooperation. Federated learning is a machine learning setting that can protect data privacy, but the high communication cost is often the bottleneck of the federated systems, especially for large neural networks. Limiting the number and size of communications is necessary for the practical training of large neural structures. Gradient sparsification has received increasing attention as a method to reduce communication cost, which only updates significant gradients and accumulates insignificant gradients locally. However, the secure aggregation framework cannot directly use gradient sparsification. This article proposes two sparsification methods to reduce communication cost in federated learning. One is a time-varying hierarchical sparsification method for model parameter update, which solves the problem of maintaining model accuracy after high ratio sparsity. It can significantly reduce the cost of a single communication. The other is to apply the sparsification method to the secure aggregation framework. We sparse the encryption mask matrix to reduce the cost of communication while protecting privacy. Experiments show that under different Non-IID experiment settings, our method can reduce the upload communication cost to about 2.9% to 18.9% of the conventional federated learning algorithm when the sparse rate is 0.01.
With the rapid advancement of 5G networks, billions of smart Internet of Things (IoT) devices along with an enormous amount of data are generated at the network edge. While still at an early age, it is expected that the evolving 6G network will adopt advanced artificial intelligence (AI) technologies to collect, transmit, and learn this valuable data for innovative applications and intelligent services. However, traditional machine learning (ML) approaches require centralizing the training data in the data center or cloud, raising serious user-privacy concerns. Federated learning, as an emerging distributed AI paradigm with privacy-preserving nature, is anticipated to be a key enabler for achieving ubiquitous AI in 6G networks. However, there are several system and statistical heterogeneity challenges for effective and efficient FL implementation in 6G networks. In this article, we investigate the optimization approaches that can effectively address the challenging heterogeneity issues from three aspects: incentive mechanism design, network resource management, and personalized model optimization. We also present some open problems and promising directions for future research.
Our research deals with the optimization version of the set partition problem, where the objective is to minimize the absolute difference between the sums of the two disjoint partitions. Although this problem is known to be NP-hard and requires exponential time to solve, we propose a less demanding version of this problem where the goal is to find a locally optimal solution. In our approach, we consider the local optimality in respect to any movement of at most two elements. To accomplish this, we developed an algorithm that can generate a locally optimal solution in at most $O(N^2)$ time and $O(N)$ space. Our algorithm can handle arbitrary input precisions and does not require positive or integer inputs. Hence, it can be applied in various problem scenarios with ease.
There are multiple cluster randomised trial designs that vary in when the clusters cross between control and intervention states, when observations are made within clusters, and how many observations are made at that time point. Identifying the most efficient study design is complex though, owing to the correlation between observations within clusters and over time. In this article, we present a review of statistical and computational methods for identifying optimal cluster randomised trial designs. We also adapt methods from the experimental design literature for experimental designs with correlated observations to the cluster trial context. We identify three broad classes of methods: using exact formulae for the treatment effect estimator variance for specific models to derive algorithms or weights for cluster sequences; generalised methods for estimating weights for experimental units; and, combinatorial optimisation algorithms to select an optimal subset of experimental units. We also discuss methods for rounding weights to whole numbers of clusters and extensions to non-Gaussian models. We present results from multiple cluster trial examples that compare the different methods, including problems involving determining optimal allocation of clusters across a set of cluster sequences, and selecting the optimal number of single observations to make in each cluster-period for both Gaussian and non-Gaussian models, and including exchangeable and exponential decay covariance structures.
Yao graphs are geometric spanners that connect each point of a given point set to its nearest neighbor in each of $k$ cones drawn around it. Yao graphs were introduced to construct minimum spanning trees in $d$ dimensional spaces. Moreover, they are used for instance in topology control in wireless networks. An optimal \Onlogn time algorithm to construct Yao graphs for given point set has been proposed in the literature but -- to the best of our knowledge -- never been implemented. Instead, algorithms with a quadratic complexity are used in popular packages to construct these graphs. In this paper we present the first implementation of the optimal Yao graph algorithm. We develop and tune the data structures required to achieve the O(n log n) bound and detail algorithmic adaptions necessary to take the original algorithm from theory to practice. We propose a priority queue data structure that separates static and dynamic events and might be of independent interest for other sweepline algorithms. Additionally, we propose a new Yao graph algorithm based on a uniform grid data structure that performs well for medium-sized inputs. We evaluate our implementations on a wide variety synthetic and real-world datasets and show that our implementation outperforms current publicly available implementations by at least an order of magnitude.
Stochastic gradient descent (SGD) is an estimation tool for large data employed in machine learning and statistics. Due to the Markovian nature of the SGD process, inference is a challenging problem. An underlying asymptotic normality of the averaged SGD (ASGD) estimator allows for the construction of a batch-means estimator of the asymptotic covariance matrix. Instead of the usual increasing batch-size strategy employed in ASGD, we propose a memory efficient equal batch-size strategy and show that under mild conditions, the estimator is consistent. A key feature of the proposed batching technique is that it allows for bias-correction of the variance, at no cost to memory. Since joint inference for high dimensional problems may be undesirable, we present marginal-friendly simultaneous confidence intervals, and show through an example how covariance estimators of ASGD can be employed in improved predictions.
Given a basic block of instructions, finding a schedule that requires the minimum number of registers for evaluation is a well-known problem. The problem is NP-complete when the dependences among instructions form a directed-acyclic graph instead of a tree. We are striving to find efficient approximation algorithms for this problem not simply because it is an interesting graph optimization problem in theory. A good solution to this problem is also an essential component in solving the more complex instruction scheduling problem on GPU. In this paper, we start with explanations on why this problem is important in GPU instruction scheduling. We then explore two different approaches to tackling this problem. First we model this problem as a constraint-programming problem. Using a state-of-the-art CP-SAT solver, we can find optimal answers for much larger cases than previous works on a modest desktop PC. Second, guided by the optimal answers, we design and evaluate heuristics that can be applied to the polynomial-time list scheduling algorithms. A combination of those heuristics can achieve the register-pressure results that are about 16\% higher than the optimal minimum on average. However, there are still near 3\% cases in which the register pressure by the heuristic approach is 50\% higher than the optimal minimum.
In recent years, implicit deep learning has emerged as a method to increase the effective depth of deep neural networks. While their training is memory-efficient, they are still significantly slower to train than their explicit counterparts. In Deep Equilibrium Models (DEQs), the training is performed as a bi-level problem, and its computational complexity is partially driven by the iterative inversion of a huge Jacobian matrix. In this paper, we propose a novel strategy to tackle this computational bottleneck from which many bi-level problems suffer. The main idea is to use the quasi-Newton matrices from the forward pass to efficiently approximate the inverse Jacobian matrix in the direction needed for the gradient computation. We provide a theorem that motivates using our method with the original forward algorithms. In addition, by modifying these forward algorithms, we further provide theoretical guarantees that our method asymptotically estimates the true implicit gradient. We empirically study this approach and the recent Jacobian-Free method in different settings, ranging from hyperparameter optimization to large Multiscale DEQs (MDEQs) applied to CIFAR and ImageNet. Both methods reduce significantly the computational cost of the backward pass. While SHINE has a clear advantage on hyperparameter optimization problems, both methods attain similar computational performances for larger scale problems such as MDEQs at the cost of a limited performance drop compared to the original models.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191