Deep Learning optimization involves minimizing a high-dimensional loss function in the weight space which is often perceived as difficult due to its inherent difficulties such as saddle points, local minima, ill-conditioning of the Hessian and limited compute resources. In this paper, we provide a comprehensive review of 12 standard optimization methods successfully used in deep learning research and a theoretical assessment of the difficulties in numerical optimization from the optimization literature.

We propose an efficient method for approximating natural gradient descent in neural networks which we call Kronecker-factored Approximate Curvature (K-FAC). K-FAC is based on an efficiently invertible approximation of a neural network's Fisher information matrix which is neither diagonal nor low-rank, and in some cases is completely non-sparse. It is derived by approximating various large blocks of the Fisher (corresponding to entire layers) as being the Kronecker product of two much smaller matrices. While only several times more expensive to compute than the plain stochastic gradient, the updates produced by K-FAC make much more progress optimizing the objective, which results in an algorithm that can be much faster than stochastic gradient descent with momentum in practice. And unlike some previously proposed approximate natural-gradient/Newton methods which use high-quality non-diagonal curvature matrices (such as Hessian-free optimization), K-FAC works very well in highly stochastic optimization regimes. This is because the cost of storing and inverting K-FAC's approximation to the curvature matrix does not depend on the amount of data used to estimate it, which is a feature typically associated only with diagonal or low-rank approximations to the curvature matrix.

We explore the usage of the Levenberg-Marquardt (LM) algorithm for regression (non-linear least squares) and classification (generalized Gauss-Newton methods) tasks in neural networks. We compare the performance of the LM method with other popular first-order algorithms such as SGD and Adam, as well as other second-order algorithms such as L-BFGS , Hessian-Free and KFAC. We further speed up the LM method by using adaptive momentum, learning rate line search, and uphill step acceptance.

多目标优化（MOO）旨在同时优化多个冲突的目标，并在机器学习中发现了重要的应用，例如最大程度地减少分类损失和差异，以在处理不同的人群方面以保持公平。最佳性，进一步优化一个目标至少将至少损害另一个目标，而决策者需要全面探索多个Optima（称为Pareto Front），以确定一个最终解决方案。我们解决了寻找帕累托阵线的效率。首先，使用随机多偏差下降（SMGD）从头开始寻找前部，对于大型神经网络和数据集很昂贵。我们建议基于预测器 - 校正方法来探索帕累托阵线作为一些初始Optima的歧管。其次，对于每个探索步骤，预测变量求解一个大规模的线性系统，该系统在模型参数数量中二次缩放，并且需要一个反向传播来评估求解器的二阶Hessian-vector产品。我们提出了一个只能线性缩放的高斯 - 纽顿近似，并且只需要每次迭代的一阶内产物。这还允许在大约求解线性系统时，在微小和共轭梯度方法之间进行选择。这些创新使大型网络成为可能的预测器 - 校准。关于多目标（公平和准确性）错误信息检测任务的实验表明，1）预测器 - 矫正器方法可以在更少的时间内找到比或与SMGD更好或与SMGD相似的方法； 2）提出的一阶方法不会损害二阶方法识别的帕累托前沿的质量，同时进一步缩短了运行时间。

Deep and recurrent neural networks (DNNs and RNNs respectively) are powerful models that were considered to be almost impossible to train using stochastic gradient descent with momentum. In this paper, we show that when stochastic gradient descent with momentum uses a well-designed random initialization and a particular type of slowly increasing schedule for the momentum parameter, it can train both DNNs and RNNs (on datasets with long-term dependencies) to levels of performance that were previously achievable only with Hessian-Free optimization. We find that both the initialization and the momentum are crucial since poorly initialized networks cannot be trained with momentum and well-initialized networks perform markedly worse when the momentum is absent or poorly tuned.Our success training these models suggests that previous attempts to train deep and recurrent neural networks from random initializations have likely failed due to poor initialization schemes. Furthermore, carefully tuned momentum methods su ce for dealing with the curvature issues in deep and recurrent network training objectives without the need for sophisticated second-order methods.

深度学习在广泛的AI应用方面取得了有希望的结果。较大的数据集和模型一致地产生更好的性能。但是，我们一般花费更长的培训时间，以更多的计算和沟通。在本调查中，我们的目标是在模型精度和模型效率方面提供关于大规模深度学习优化的清晰草图。我们调查最常用于优化的算法，详细阐述了大批量培训中出现的泛化差距的可辩论主题，并审查了解决通信开销并减少内存足迹的SOTA策略。

We propose a Randomised Subspace Gauss-Newton (R-SGN) algorithm for solving nonlinear least-squares optimization problems, that uses a sketched Jacobian of the residual in the variable domain and solves a reduced linear least-squares on each iteration. A sublinear global rate of convergence result is presented for a trust-region variant of R-SGN, with high probability, which matches deterministic counterpart results in the order of the accuracy tolerance. Promising preliminary numerical results are presented for R-SGN on logistic regression and on nonlinear regression problems from the CUTEst collection.

A core capability of intelligent systems is the ability to quickly learn new tasks by drawing on prior experience. Gradient (or optimization) based meta-learning has recently emerged as an effective approach for few-shot learning. In this formulation, meta-parameters are learned in the outer loop, while task-specific models are learned in the inner-loop, by using only a small amount of data from the current task. A key challenge in scaling these approaches is the need to differentiate through the inner loop learning process, which can impose considerable computational and memory burdens. By drawing upon implicit differentiation, we develop the implicit MAML algorithm, which depends only on the solution to the inner level optimization and not the path taken by the inner loop optimizer. This effectively decouples the meta-gradient computation from the choice of inner loop optimizer. As a result, our approach is agnostic to the choice of inner loop optimizer and can gracefully handle many gradient steps without vanishing gradients or memory constraints. Theoretically, we prove that implicit MAML can compute accurate meta-gradients with a memory footprint no more than that which is required to compute a single inner loop gradient and at no overall increase in the total computational cost. Experimentally, we show that these benefits of implicit MAML translate into empirical gains on few-shot image recognition benchmarks.

目前，深层神经网络（DNN）主要使用一阶方法进行训练。其中一些方法（例如Adam，Adagrad和Rmsprop及其变体）通过使用对角线矩阵来预先处理随机梯度。最近，通过通过按层块 - diagonal矩阵对随机梯度进行预处理，已开发出有效的二阶方法，例如KFAC，K-BFGS，洗发水和TNT。在这里，我们提出了一种自适应的“迷你块Fisher（MBF）”预处理方法，其中在这两类方法之间。具体而言，我们的方法对经验渔民矩阵使用块对基近似值，在DNN中的每一层（无论是卷积还是馈送）和完全连接，相关的对角线本身都是块 - diagonal，并且由A组成。大量适度的迷你块。我们的新方法利用GPU的并行性来有效地对每一层的大量矩阵进行计算。因此，MBF的均值计算成本仅略高于一阶方法。将我们提出的方法的性能与在自动编码器和CNN问题上的几种基线方法进行了比较，以在时间效率和概括功率方面验证其有效性。最后，证明MBF的理想化版本线性收敛。

我们研究了使用尖刺，现场依赖的随机矩阵理论研究迷你批次对深神经网络损失景观的影响。我们表明，批量黑森州的极值值的大小大于经验丰富的黑森州。我们还获得了类似的结果对Hessian的概括高斯牛顿矩阵近似。由于我们的定理，我们推导出作为批量大小的最大学习速率的分析表达式，为随机梯度下降（线性缩放）和自适应算法（例如ADAM（Square Root Scaling）提供了通知实际培训方案，例如光滑，非凸深神经网络。虽然随机梯度下降的线性缩放是在我们概括的更多限制性条件下导出的，但是适应优化者的平方根缩放规则是我们的知识，完全小说。随机二阶方法和自适应方法的百分比，我们得出了最小阻尼系数与学习率与批量尺寸的比率成比例。我们在Cifar-$ 100 $和ImageNet数据集上验证了我们的VGG / WimerEsnet架构上的索赔。根据我们对象检的调查，我们基于飞行学习率和动量学习者开发了一个随机兰齐齐竞争，这避免了对这些关键的超参数进行昂贵的多重评估的需求，并在预残留的情况下显示出良好的初步结果Cifar的architecure - $ 100 $。

二阶优化器被认为具有加快神经网络训练的潜力，但是由于曲率矩阵的尺寸巨大，它们通常需要近似值才能计算。最成功的近似家庭是Kronecker因块状曲率估计值（KFAC）。在这里，我们结合了先前工作的工具，以评估确切的二阶更新和仔细消融以建立令人惊讶的结果：由于其近似值，KFAC与二阶更新无关，尤其是，它极大地胜过真实的第二阶段更新。订单更新。这一挑战广泛地相信，并立即提出了为什么KFAC表现如此出色的问题。为了回答这个问题，我们提出了强烈的证据，表明KFAC近似于一阶算法，该算法在神经元上执行梯度下降而不是权重。最后，我们表明，这种优化器通常会在计算成本和数据效率方面改善KFAC。

培训深度神经网络消耗了许多计算中心的计算资源份额。通常，采用蛮力的方法来获得高参数值。我们的目标是（1）通过启用对大型神经网络的二阶优化方法来增强此功能，以及（2）对特定任务进行性能优化器进行调查，以建议用户最适合他们的问题。我们介绍了一种新颖的二阶优化方法，该方法仅需要Hessian对向量的影响，并避免明确设置大型网络的Hessian的巨大成本。我们将提出的二阶方法与两个最先进的优化器进行了比较，这些方法在五个代表性的神经网络问题上进行了比较，包括回归和来自计算机视觉或变异自动编码器的非常深的网络。对于最大的设置，我们将优化器与HOROVOD有效平行，并将其应用于8 GPU NVIDIA P100（DGX-1）机器。

目前的论文研究了最小化损失$ f（\ boldsymbol {x}）$的问题，而在s $ \ boldsymbol {d} \ boldsymbol {x} \的约束，其中$ s $是一个关闭的集合，凸面或非，$ \ boldsymbol {d} $是熔化参数的矩阵。融合约束可以捕获平滑度，稀疏或更一般的约束模式。为了解决这个通用的问题，我们将Beltrami-Courant罚球方法与近距离原则相结合。后者是通过最小化惩罚目标的推动$ f（\ boldsymbol {x}）+ \ frac {\ rho} {2} \ text {dist}（\ boldsymbol {d} \ boldsymbol {x}，s）^ 2 $涉及大型调整常量$ \ rho $和$ \ boldsymbol {d} \ boldsymbol {x} $的平方欧几里德距离$ s $。通过最小化大多数代理函数$ f（\ boldsymbol {x}，从当前迭代$ \ boldsymbol {x} _n $构建相应的近距离算法的下一个迭代$ \ boldsymbol {x} _ {n + 1} $。 ）+ \ frac {\ rho} {2} \ | \ boldsymbol {d} \ boldsymbol {x} - \ mathcal {p} _ {s}（\ boldsymbol {d} \ boldsymbol {x} _n）\ | ^ 2 $。对于固定$ \ rho $和subanalytic损失$ f（\ boldsymbol {x}）$和子质约束设置$ s $，我们证明了汇聚点。在更强大的假设下，我们提供了收敛速率并展示线性本地收敛性。我们还构造了一个最陡的下降（SD）变型，以避免昂贵的线性系统解决。为了基准我们的算法，我们比较乘法器（ADMM）的交替方向方法。我们广泛的数值测试包括在度量投影，凸回归，凸聚类，总变化图像去噪和矩阵的投影到良好状态数的问题。这些实验表明了我们在高维问题上最陡的速度和可接受的准确性。

Many problems in machine learning involve bilevel optimization (BLO), including hyperparameter optimization, meta-learning, and dataset distillation. Bilevel problems consist of two nested sub-problems, called the outer and inner problems, respectively. In practice, often at least one of these sub-problems is overparameterized. In this case, there are many ways to choose among optima that achieve equivalent objective values. Inspired by recent studies of the implicit bias induced by optimization algorithms in single-level optimization, we investigate the implicit bias of gradient-based algorithms for bilevel optimization. We delineate two standard BLO methods -- cold-start and warm-start -- and show that the converged solution or long-run behavior depends to a large degree on these and other algorithmic choices, such as the hypergradient approximation. We also show that the inner solutions obtained by warm-start BLO can encode a surprising amount of information about the outer objective, even when the outer parameters are low-dimensional. We believe that implicit bias deserves as central a role in the study of bilevel optimization as it has attained in the study of single-level neural net optimization.

我们引入了一种降低尺寸的二阶方法（DRSOM），用于凸和非凸的不受约束优化。在类似信任区域的框架下，我们的方法保留了二阶方法的收敛性，同时仅在两个方向上使用Hessian-Vector产品。此外，计算开销仍然与一阶相当，例如梯度下降方法。我们证明该方法的复杂性为$ O（\ epsilon^{ - 3/2}）$，以满足子空间中的一阶和二阶条件。DRSOM的适用性和性能通过逻辑回归，$ L_2-L_P $最小化，传感器网络定位和神经网络培训的各种计算实验展示。对于神经网络，我们的初步实施似乎在训练准确性和迭代复杂性方面与包括SGD和ADAM在内的最先进的一阶方法获得了计算优势。

牛顿方法和Adagrad等高级优化算法受益于二阶导数或二阶统计，以实现更好的下降方向和更快的收敛速率。在他们的心中，这种算法需要计算矩阵的矩阵的反平方根或反平方根，其大小是搜索空间维度的二次。对于高维搜索空间，平方根的矩阵反转或反转变为压倒性的，进而需要近似方法。在这项工作中，我们提出了一种新的矩阵近似方法，该方法将矩阵分为块，并将每个块代表一个或两个数字。该方法允许有效地计算矩阵逆和逆平方根。我们将我们的方法应用于Adagrad，以培训深层神经网络。实验表明与对角线近似相比令人鼓舞的结果。

我们介绍了一种牛顿型方法，可以从任何初始化和带有Lipschitz Hessians的任意凸面目标收敛。通过将立方规范化与某种自适应levenberg - Marquardt罚款合并来实现这一目标。特别地，我们表明由$ x ^ {k + 1} = x ^ k - \ bigl（\ nabla ^ 2 f（x ^ k）+ \ sqrt {h \ | \ nabla f（x ^ k）给出的迭代）\ |} \ mathbf {i} \ bigr）^ { - 1} \ nabla f（x ^ k）$，其中$ h> 0 $是一个常数，用$ \ mathcal {o}全球收敛（\ frac{1} {k ^ 2}）$率。我们的方法是牛顿方法的第一个变体，具有廉价迭代和可怕的全球融合。此外，我们证明当目的强烈凸起时，本地我们的方法会收敛超连续。为了提高方法的性能，我们提供了一种不需要超参数的线路搜索程序，并且可提供高效。

A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance. This work extends the results of .

A matrix free and a low rank approximation preconditioner are proposed to accelerate the convergence of stochastic gradient descent (SGD) by exploiting curvature information sampled from Hessian-vector products or finite differences of parameters and gradients similar to the BFGS algorithm. Both preconditioners are fitted with an online updating manner minimizing a criterion that is free of line search and robust to stochastic gradient noise, and further constrained to be on certain connected Lie groups to preserve their corresponding symmetry or invariance, e.g., orientation of coordinates by the connected general linear group with positive determinants. The Lie group's equivariance property facilitates preconditioner fitting, and its invariance property saves any need of damping, which is common in second-order optimizers, but difficult to tune. The learning rate for parameter updating and step size for preconditioner fitting are naturally normalized, and their default values work well in most situations.