晶体和分子感兴趣的特性,例如带隙,弹性和溶解度,通常相互关联:它们受相同的基础物理定律的控制。但是,当最新的图形神经网络尝试同时预测多个属性(多任务学习(MTL)设置)时,它们经常表现不佳。这表明图形网络可能无法完全利用这些潜在的相似性。在这里,我们研究了这种现象的潜在解释:每个物业损失表面的曲率都有很大变化,导致学习效率低下。曲率上的这种差异可以通过查看每个属性损耗函数的Hessians的光谱特性来评估,这是通过随机数值线性代数以无基质方式完成的。我们在两个基准数据集(材料项目(MP)和QM8)上评估我们的假设,并考虑这些发现如何为新颖的多任务学习模型的培训提供信息。
translated by 谷歌翻译
我们调查人工神经网络的损失表面Hessians的局部光谱统计数据,在那里我们发现跨多个网络架构和数据集的高斯正交集合统计数据非常一致。这些结果阐述了随机矩阵理论对神经网络建模的适用性,并提出了在深度学习中损失表面研究中的先前未被识别的作用。通过这些观察的启发,我们提出了一种新颖的神经网络的真正损失表面模型,与我们的观察结果一致,这允许Hessian光谱密度在实践中广泛观察到具有秩的退化性和异常值,并预测损失梯度的独立性越来越长重量空间中距离的函数。我们进一步调查了神经网络中真正损失表面的重要性,并与以前的工作相比,找到了定位全球最小值的指数硬度对实现最新性能的实际后果。
translated by 谷歌翻译
Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels.Recently, large scale quantum chemistry calculation and molecular dynamics simulations coupled with advances in high throughput experiments have begun to generate data at an unprecedented rate. Most classical techniques do not make effective use of the larger amounts of data that are now available. The time is ripe to apply more powerful and flexible machine learning methods to these problems, assuming we can find models with suitable inductive biases. The symmetries of atomic systems suggest neural networks that operate on graph structured data and are invariant to graph isomorphism might also be appropriate for molecules. Sufficiently successful models could someday help automate challenging chemical search problems in drug discovery or materials science.In this paper, our goal is to demonstrate effective machine learning models for chemical prediction problems
translated by 谷歌翻译
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 $。
translated by 谷歌翻译
Advancements in neural machinery have led to a wide range of algorithmic solutions for molecular property prediction. Two classes of models in particular have yielded promising results: neural networks applied to computed molecular fingerprints or expert-crafted descriptors, and graph convolutional neural networks that construct a learned molecular representation by operating on the graph structure of the molecule.However, recent literature has yet to clearly determine which of these two methods is superior when generalizing to new chemical space. Furthermore, prior research has
translated by 谷歌翻译
Graph classification is an important area in both modern research and industry. Multiple applications, especially in chemistry and novel drug discovery, encourage rapid development of machine learning models in this area. To keep up with the pace of new research, proper experimental design, fair evaluation, and independent benchmarks are essential. Design of strong baselines is an indispensable element of such works. In this thesis, we explore multiple approaches to graph classification. We focus on Graph Neural Networks (GNNs), which emerged as a de facto standard deep learning technique for graph representation learning. Classical approaches, such as graph descriptors and molecular fingerprints, are also addressed. We design fair evaluation experimental protocol and choose proper datasets collection. This allows us to perform numerous experiments and rigorously analyze modern approaches. We arrive to many conclusions, which shed new light on performance and quality of novel algorithms. We investigate application of Jumping Knowledge GNN architecture to graph classification, which proves to be an efficient tool for improving base graph neural network architectures. Multiple improvements to baseline models are also proposed and experimentally verified, which constitutes an important contribution to the field of fair model comparison.
translated by 谷歌翻译
分析高维损失函数的几何特性,例如局部曲率以及围绕损失空间某个特定点的其他Optima的存在,可以帮助您更好地理解神经网络结构,实现属性和学习绩效之间的相互作用。在这项工作中,我们将概念从高维概率和差异几何形状结合在一起,以研究低维损耗表示中的曲率特性如何取决于原始损失空间中的曲率。我们表明,如果使用随机投影,则很少在较低维表示中正确识别原始空间中的鞍点。在这样的预测中,较低维表示中的预期曲率与原始损耗空间中的平均曲率成正比。因此,原始损耗空间中的平均曲率决定了鞍点是否平均显示为最小值,最大值或几乎平坦的区域。我们使用预期曲率和平均曲率(即标准化的Hessian Trace)之间的连接来估计黑森的痕迹,而无需像Hutchinson的方法一样计算Hessian或Hessian-Vector产品。由于随机预测无法正确识别马鞍信息,因此我们建议沿着与最大和最小的主要曲线相关的Hessian指示进行预测。我们将发现与正在进行的有关损失景观平坦性和普遍性的辩论联系起来。最后,我们在不同图像分类器上的数值实验中说明了我们的方法,最高$ 7 \ times 10^6 $参数。
translated by 谷歌翻译
分子特性预测是与关键现实影响的深度学习的增长最快的应用之一。包括3D分子结构作为学习模型的输入可以提高它们对许多分子任务的性能。但是,此信息是不可行的,可以以几个现实世界应用程序所需的规模计算。我们建议预先训练模型,以推理仅给予其仅为2D分子图的分子的几何形状。使用来自自我监督学习的方法,我们最大化3D汇总向量和图形神经网络(GNN)的表示之间的相互信息,使得它们包含潜在的3D信息。在具有未知几何形状的分子上进行微调期间,GNN仍然产生隐式3D信息,并可以使用它来改善下游任务。我们表明3D预训练为广泛的性质提供了显着的改进,例如八个量子力学性能的22%的平均MAE。此外,可以在不同分子空间中的数据集之间有效地传送所学习的表示。
translated by 谷歌翻译
电子密度$ \ rho(\ vec {r})$是用密度泛函理论(dft)计算地面能量的基本变量。除了总能量之外,$ \ rho(\ vec {r})$分布和$ \ rho(\ vec {r})$的功能通常用于捕获电子规模以功能材料和分子中的关键物理化学现象。方法提供对$ \ rho(\ vec {r})的可紊乱系统,其具有少量计算成本的复杂无序系统可以是对材料相位空间的加快探索朝向具有更好功能的新材料的逆设计的游戏更换者。我们为预测$ \ rho(\ vec {r})$。该模型基于成本图形神经网络,并且在作为消息传递图的一部分的特殊查询点顶点上预测了电子密度,但仅接收消息。该模型在多个数据组中进行测试,分子(QM9),液体乙烯碳酸酯电解质(EC)和Lixniymnzco(1-Y-Z)O 2锂离子电池阴极(NMC)。对于QM9分子,所提出的模型的准确性超过了从DFT获得的$ \ Rho(\ vec {r})$中的典型变异性,以不同的交换相关功能,并显示超出最先进的准确性。混合氧化物(NMC)和电解质(EC)数据集更好的精度甚至更好。线性缩放模型同时探测成千上万点的能力允许计算$ \ Rho(\ vec {r})$的大型复杂系统,比DFT快于允许筛选无序的功能材料。
translated by 谷歌翻译
Pre-publication draft of a book to be published byMorgan & Claypool publishers. Unedited version released with permission. All relevant copyrights held by the author and publisher extend to this pre-publication draft.
translated by 谷歌翻译
在三维分子结构上运行的计算方法有可能解决生物学和化学的重要问题。特别地,深度神经网络的重视,但它们在生物分子结构域中的广泛采用受到缺乏系统性能基准或统一工具包的限制,用于与分子数据相互作用。为了解决这个问题,我们呈现Atom3D,这是一个新颖的和现有的基准数据集的集合,跨越几个密钥的生物分子。我们为这些任务中的每一个实施多种三维分子学习方法,并表明它们始终如一地提高了基于单维和二维表示的方法的性能。结构的具体选择对于性能至关重要,具有涉及复杂几何形状的任务的三维卷积网络,在需要详细位置信息的系统中表现出良好的图形网络,以及最近开发的设备越多的网络显示出显着承诺。我们的结果表明,许多分子问题符合三维分子学习的增益,并且有可能改善许多仍然过分曝光的任务。为了降低进入并促进现场进一步发展的障碍,我们还提供了一套全面的DataSet处理,模型培训和在我们的开源ATOM3D Python包中的评估工具套件。所有数据集都可以从https://www.atom3d.ai下载。
translated by 谷歌翻译
我们在点云数据上引入了一种新的局部曲率量度,称为扩散曲率。我们的措施使用扩散图的框架,包括数据扩散操作员,结构点云数据,并根据从数据的点或区域开始的随机步行的懒惰定义局部曲率。我们表明,这种懒惰直接与Riemannian几何形状的体积比较结果有关。然后,我们使用基于点云数据扩散图的神经网络估计将此标量曲率概念扩展到整个二次形式。我们展示了关于玩具数据,单细胞数据以及估计神经网络损失景观本地Hessian矩阵的应用。
translated by 谷歌翻译
The accurate prediction of physicochemical properties of chemical compounds in mixtures (such as the activity coefficient at infinite dilution $\gamma_{ij}^\infty$) is essential for developing novel and more sustainable chemical processes. In this work, we analyze the performance of previously-proposed GNN-based models for the prediction of $\gamma_{ij}^\infty$, and compare them with several mechanistic models in a series of 9 isothermal studies. Moreover, we develop the Gibbs-Helmholtz Graph Neural Network (GH-GNN) model for predicting $\ln \gamma_{ij}^\infty$ of molecular systems at different temperatures. Our method combines the simplicity of a Gibbs-Helmholtz-derived expression with a series of graph neural networks that incorporate explicit molecular and intermolecular descriptors for capturing dispersion and hydrogen bonding effects. We have trained this model using experimentally determined $\ln \gamma_{ij}^\infty$ data of 40,219 binary-systems involving 1032 solutes and 866 solvents, overall showing superior performance compared to the popular UNIFAC-Dortmund model. We analyze the performance of GH-GNN for continuous and discrete inter/extrapolation and give indications for the model's applicability domain and expected accuracy. In general, GH-GNN is able to produce accurate predictions for extrapolated binary-systems if at least 25 systems with the same combination of solute-solvent chemical classes are contained in the training set and a similarity indicator above 0.35 is also present. This model and its applicability domain recommendations have been made open-source at https://github.com/edgarsmdn/GH-GNN.
translated by 谷歌翻译
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 .
translated by 谷歌翻译
计算催化和机器学习社区在开发用于催化剂发现和设计的机器学习模型方面取得了长足的进步。然而,跨越催化的化学空间的一般机器学习潜力仍然无法触及。一个重大障碍是在广泛的材料中获得访问培训数据的访问。缺乏数据的一类重要材料是氧化物,它抑制模型无法更广泛地研究氧气进化反应和氧化物电催化。为了解决这个问题,我们开发了开放的催化剂2022(OC22)数据集,包括62,521个密度功能理论(DFT)放松(〜9,884,504个单点计算),遍及一系列氧化物材料,覆盖范围,覆盖率和吸附物( *H, *o, *o, *o, *o, *o, * n, *c, *ooh, *oh, *oh2, *o2, *co)。我们定义广义任务,以预测催化过程中适用的总系统能量,发展几个图神经网络的基线性能(Schnet,Dimenet ++,Forcenet,Spinconv,Painn,Painn,Gemnet-DT,Gemnet-DT,Gemnet-OC),并提供预先定义的数据集分割以建立明确的基准,以实现未来的努力。对于所有任务,我们研究组合数据集是否会带来更好的结果,即使它们包含不同的材料或吸附物。具体而言,我们在Open Catalyst 2020(OC20)数据集和OC22上共同训练模型,或OC22上的微调OC20型号。在最一般的任务中,Gemnet-OC看到通过微调来提高了约32%的能量预测,通过联合训练的力预测提高了约9%。令人惊讶的是,OC20和较小的OC22数据集的联合培训也将OC20的总能量预测提高了约19%。数据集和基线模型是开源的,公众排行榜将遵循,以鼓励社区的持续发展,以了解总能源任务和数据。
translated by 谷歌翻译
本文考虑了深神经网络中随机矩阵普遍性的几个方面。在最近的实验工作中,我们使用与局部统计相关的随机矩阵的普遍特性,以基于其Hessians的现实模型来获得对深神经网络的实际含义。特别是,我们得出了深度神经网络光谱中异常值的普遍方面,并证明了随机矩阵局部定律在流行的预处理梯度下降算法中的重要作用。我们还通过基于统计物理学和随机矩阵理论的工具的一般参数,对深度神经网络损失表面的见解。
translated by 谷歌翻译
可拍照的分子显示了可以使用光访问的两个或多个异构体形式。将这些异构体的电子吸收带分开是选择性解决特定异构体并达到高光稳态状态的关键,同时总体红色转移带来的吸收带可以限制因紫外线暴露而限制材料损害,并增加了光疗法应用中的渗透深度。但是,通过合成设计将这些属性工程为系统仍然是一个挑战。在这里,我们提出了一条数据驱动的发现管道,用于由数据集策划和使用高斯过程的多任务学习支撑的分子照片开关。在对电子过渡波长的预测中,我们证明了使用来自四个Photoswitch转变波长的标签训练的多输出高斯过程(MOGP)产生相对于单任务模型的最强预测性能,并且在操作上超过了时间依赖时间依赖性的密度理论(TD) -dft)就预测的墙壁锁定时间而言。我们通过筛选可商购的可拍摄分子库来实验验证我们提出的方法。通过此屏幕,我们确定了几个图案,这些基序显示了它们的异构体的分离电子吸收带,表现出红移的吸收,并且适用于信息传输和光电学应用。我们的策划数据集,代码以及所有型号均可在https://github.com/ryan-rhys/the-photoswitch-dataset上提供
translated by 谷歌翻译
物理信息的神经网络(PINN)是神经网络(NNS),它们作为神经网络本身的组成部分编码模型方程,例如部分微分方程(PDE)。如今,PINN是用于求解PDE,分数方程,积分分化方程和随机PDE的。这种新颖的方法已成为一个多任务学习框架,在该框架中,NN必须在减少PDE残差的同时拟合观察到的数据。本文对PINNS的文献进行了全面的综述:虽然该研究的主要目标是表征这些网络及其相关的优势和缺点。该综述还试图将出版物纳入更广泛的基于搭配的物理知识的神经网络,这些神经网络构成了香草·皮恩(Vanilla Pinn)以及许多其他变体,例如物理受限的神经网络(PCNN),各种HP-VPINN,变量HP-VPINN,VPINN,VPINN,变体。和保守的Pinn(CPINN)。该研究表明,大多数研究都集中在通过不同的激活功能,梯度优化技术,神经网络结构和损耗功能结构来定制PINN。尽管使用PINN的应用范围广泛,但通过证明其在某些情况下比有限元方法(FEM)等经典数值技术更可行的能力,但仍有可能的进步,最著名的是尚未解决的理论问题。
translated by 谷歌翻译
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.
translated by 谷歌翻译
近年来,基于Weisfeiler-Leman算法的算法和神经架构,是一个众所周知的Graph同构问题的启发式问题,它成为具有图形和关系数据的机器学习的强大工具。在这里,我们全面概述了机器学习设置中的算法的使用,专注于监督的制度。我们讨论了理论背景,展示了如何将其用于监督的图形和节点表示学习,讨论最近的扩展,并概述算法的连接(置换 - )方面的神经结构。此外,我们概述了当前的应用和未来方向,以刺激进一步的研究。
translated by 谷歌翻译