A Neural Network Gaussian Process (NNGP) is a Gaussian process (GP) obtained as the limit of a certain type of sequence of neural networks. Specifically, a wide variety of network architectures converges to a GP in the infinitely wide limit, in the sense of distribution. The concept constitutes an intensional definition, i.e., a NNGP is just a GP, but distinguished by how it is obtained. == Motivation == Bayesian networks are a modeling tool for assigning probabilities to events, and thereby characterizing the uncertainty in a model's predictions. Deep learning and artificial neural networks are approaches used in machine learning to build computational models which learn from training examples. Bayesian neural networks merge these fields. They are a type of neural network whose parameters and predictions are both probabilistic. While standard neural networks often assign high confidence even to incorrect predictions, Bayesian neural networks can more accurately evaluate how likely their predictions are to be correct. Computation in artificial neural networks is usually organized into sequential layers of artificial neurons. The number of neurons in a layer is called the layer width. When we consider a sequence of Bayesian neural networks with increasingly wide layers (see figure), they converge in distribution to a NNGP. This large width limit is of practical interest, since the networks often improve as layers get wider. And the process may give a closed form way to evaluate networks. NNGPs also appears in several other contexts: It describes the distribution over predictions made by wide non-Bayesian artificial neural networks after random initialization of their parameters, but before training; it appears as a term in neural tangent kernel prediction equations; it is used in deep information propagation to characterize whether hyperparameters and architectures will be trainable. It is related to other large width limits of neural networks. === Scope === The first correspondence result had been established in the 1995 PhD thesis of Radford M. Neal, then supervised by Geoffrey Hinton at University of Toronto. Neal cites David J. C. MacKay as inspiration, who worked in Bayesian learning. Today the correspondence is proven for: Single hidden layer Bayesian neural networks; deep fully connected networks as the number of units per layer is taken to infinity; convolutional neural networks as the number of channels is taken to infinity; transformer networks as the number of attention heads is taken to infinity; recurrent networks as the number of units is taken to infinity. In fact, this NNGP correspondence holds for almost any architecture: Generally, if an architecture can be expressed solely via matrix multiplication and coordinatewise nonlinearities (i.e., a tensor program), then it has an infinite-width GP. This in particular includes all feedforward or recurrent neural networks composed of multilayer perceptron, recurrent neural networks (e.g., LSTMs, GRUs), (nD or graph) convolution, pooling, skip connection, attention, batch normalization, and/or layer normalization. === Illustration === Every setting of a neural network's parameters θ {\displaystyle \theta } corresponds to a specific function computed by the neural network. A prior distribution p ( θ ) {\displaystyle p(\theta )} over neural network parameters therefore corresponds to a prior distribution over functions computed by the network. As neural networks are made infinitely wide, this distribution over functions converges to a Gaussian process for many architectures. The notation used in this section is the same as the notation used below to derive the correspondence between NNGPs and fully connected networks, and more details can be found there. The figure to the right plots the one-dimensional outputs z L ( ⋅ ; θ ) {\displaystyle z^{L}(\cdot ;\theta )} of a neural network for two inputs x {\displaystyle x} and x ∗ {\displaystyle x^{}} against each other. The black dots show the function computed by the neural network on these inputs for random draws of the parameters from p ( θ ) {\displaystyle p(\theta )} . The red lines are iso-probability contours for the joint distribution over network outputs z L ( x ; θ ) {\displaystyle z^{L}(x;\theta )} and z L ( x ∗ ; θ ) {\displaystyle z^{L}(x^{};\theta )} induced by p ( θ ) {\displaystyle p(\theta )} . This is the distribution in function space corresponding to the distribution p ( θ ) {\displaystyle p(\theta )} in parameter space, and the black dots are samples from this distribution. For infinitely wide neural networks, since the distribution over functions computed by the neural network is a Gaussian process, the joint distribution over network outputs is a multivariate Gaussian for any finite set of network inputs. == Discussion == === Infinitely wide fully connected network === This section expands on the correspondence between infinitely wide neural networks and Gaussian processes for the specific case of a fully connected architecture. It provides a proof sketch outlining why the correspondence holds, and introduces the specific functional form of the NNGP for fully connected networks. The proof sketch closely follows the approach by Novak and coauthors. ==== Network architecture specification ==== Consider a fully connected artificial neural network with inputs x {\displaystyle x} , parameters θ {\displaystyle \theta } consisting of weights W l {\displaystyle W^{l}} and biases b l {\displaystyle b^{l}} for each layer l {\displaystyle l} in the network, pre-activations (pre-nonlinearity) z l {\displaystyle z^{l}} , activations (post-nonlinearity) y l {\displaystyle y^{l}} , pointwise nonlinearity ϕ ( ⋅ ) {\displaystyle \phi (\cdot )} , and layer widths n l {\displaystyle n^{l}} . For simplicity, the width n L + 1 {\displaystyle n^{L+1}} of the readout vector z L {\displaystyle z^{L}} is taken to be 1. The parameters of this network have a prior distribution p ( θ ) {\displaystyle p(\theta )} , which consists of an isotropic Gaussian for each weight and bias, with the variance of the weights scaled inversely with layer width. This network is illustrated in the figure to the right, and described by the following set of equations: x ≡ input y l ( x ) = { x l = 0 ϕ ( z l − 1 ( x ) ) l > 0 z i l ( x ) = ∑ j W i j l y j l ( x ) + b i l W i j l ∼ N ( 0 , σ w 2 n l ) b i l ∼ N ( 0 , σ b 2 ) ϕ ( ⋅ ) ≡ nonlinearity y l ( x ) , z l − 1 ( x ) ∈ R n l × 1 n L + 1 = 1 θ = { W 0 , b 0 , … , W L , b L } {\displaystyle {\begin{aligned}x&\equiv {\text{input}}\\y^{l}(x)&=\left\{{\begin{array}{lcl}x&&l=0\\\phi \left(z^{l-1}(x)\right)&&l>0\end{array}}\right.\\z_{i}^{l}(x)&=\sum _{j}W_{ij}^{l}y_{j}^{l}(x)+b_{i}^{l}\\W_{ij}^{l}&\sim {\mathcal {N}}\left(0,{\frac {\sigma _{w}^{2}}{n^{l}}}\right)\\b_{i}^{l}&\sim {\mathcal {N}}\left(0,\sigma _{b}^{2}\right)\\\phi (\cdot )&\equiv {\text{nonlinearity}}\\y^{l}(x),z^{l-1}(x)&\in \mathbb {R} ^{n^{l}\times 1}\\n^{L+1}&=1\\\theta &=\left\{W^{0},b^{0},\dots ,W^{L},b^{L}\right\}\end{aligned}}} ==== ==== z l | y l {\displaystyle z^{l}|y^{l}} is a Gaussian process We first observe that the pre-activations z l {\displaystyle z^{l}} are described by a Gaussian process conditioned on the preceding activations y l {\displaystyle y^{l}} . This result holds even at finite width. Each pre-activation z i l {\displaystyle z_{i}^{l}} is a weighted sum of Gaussian random variables, corresponding to the weights W i j l {\displaystyle W_{ij}^{l}} and biases b i l {\displaystyle b_{i}^{l}} , where the coefficients for each of those Gaussian variables are the preceding activations y j l {\displaystyle y_{j}^{l}} . Because they are a weighted sum of zero-mean Gaussians, the z i l {\displaystyle z_{i}^{l}} are themselves zero-mean Gaussians (conditioned on the coefficients y j l {\displaystyle y_{j}^{l}} ). Since the z l {\displaystyle z^{l}} are jointly Gaussian for any set of y l {\displaystyle y^{l}} , they are described by a Gaussian process conditioned on the preceding activations y l {\displaystyle y^{l}} . The covariance or kernel of this Gaussian process depends on the weight and bias variances σ w 2 {\displaystyle \sigma _{w}^{2}} and σ b 2 {\displaystyle \sigma _{b}^{2}} , as well as the second moment matrix K l {\displaystyle K^{l}} of the preceding activations y l {\displaystyle y^{l}} , z i l ∣ y l ∼ G P ( 0 , σ w 2 K l + σ b 2 ) K l ( x , x ′ ) = 1 n l ∑ i y i l ( x ) y i l ( x ′ ) {\displaystyle {\begin{aligned}z_{i}^{l}\mid y^{l}&\sim {\mathcal {GP}}\left(0,\sigma _{w}^{2}K^{l}+\sigma _{b}^{2}\right)\\K^{l}(x,x')&={\frac {1}{n^{l}}}\sum _{i}y_{i}^{l}(x)y_{i}^{l}(x')\end{aligned}}} The effect of the weight scale σ w 2 {\displaystyle \sigma _{w}^{2}} is to rescale the contribution to the covariance matrix from K l {\displaystyle K^{l}} , while the bias is shared for all inputs, and so σ b 2 {\displaystyle \sigma _{b}^{2}} makes the z i l {\displaystyle z_{i}^{l}} for different datapoints more similar and
Multi-focus image fusion
Multi-focus image fusion is a multiple image compression technique using input images with different focus depths to make one output image that preserves all information. == Overview == The main idea of image fusion is gathering important and the essential information from the input images into one single image which ideally has all of the information of the input images. The research history of image fusion spans over 30 years and many scientific papers. Image fusion generally has two aspects: image fusion methods and objective evaluation metrics. In visual sensor networks (VSN), sensors are cameras which record images and video sequences. In many applications of VSN, a camera can't give a perfect illustration including all details of the scene. This is because of the limited depth of focus of the optical lens of cameras. Therefore, just the object located in the focal length of camera is focused and clear, and other parts of the image are blurred. VSN captures images with different depths of focus using several cameras. Due to the large amount of data generated by cameras compared to other sensors such as pressure and temperature sensors and some limitations of bandwidth, energy consumption and processing time, it is essential to process the local input images to decrease the amount of transmitted data. == Multi-Focus image fusion in the spatial domain == Huang and Jing have reviewed and applied several focus measurements in the spatial domain for the multi-focus image fusion process, suitable for real-time applications. They mentioned some focus measurements including variance, energy of image gradient (EOG), Tenenbaum's algorithm (Tenengrad), energy of Laplacian (EOL), sum-modified-Laplacian (SML), and spatial frequency (SF). Their experiments showed that EOL gave better results than other methods like variance and spatial frequency. == Multi-Focus image fusion in multi-scale transform and DCT domain == Image fusion based on the multi-scale transform is the most commonly used and promising technique. Laplacian pyramid transform, gradient pyramid-based transform, morphological pyramid transform and the premier ones, discrete wavelet transform, shift-invariant wavelet transform (SIDWT), and discrete cosine harmonic wavelet transform (DCHWT) are some examples of image fusion methods based on multi-scale transform. These methods are complex and have some limitations e.g. processing time and energy consumption. For example, multi-focus image fusion methods based on DWT require a lot of convolution operations, so they take more time and energy to process. Therefore, most methods in multi-scale transform are not suitable for real-time applications. Moreover, these methods are not very successful along edges, due to the wavelet transform process missing the edges of the image. They create ringing artefacts in the output image and reduce its quality. Due to the aforementioned problems in the multi-scale transform methods, researchers are interested in multi-focus image fusion in the DCT domain. DCT-based methods are more efficient in terms of transmission and archiving images coded in Joint Photographic Experts Group (JPEG) standard to the upper node in the VSN agent. A JPEG system consists of a pair of an encoder and a decoder. In the encoder, images are divided into non-overlapping 8×8 blocks, and the DCT coefficients are calculated for each. Since the quantization of DCT coefficients is a lossy process, many of the small-valued DCT coefficients are quantized to zero, which corresponds to high frequencies. DCT-based image fusion algorithms work better when the multi-focus image fusion methods are applied in the compressed domain. In addition, in the spatial-based methods, the input images must be decoded and then transferred to the spatial domain. After implementation of the image fusion operations, the output fused images must again be encoded. DCT domain-based methods do not require complex and time-consuming consecutive decoding and encoding operations. Therefore, the image fusion methods based on DCT domain operate with much less energy and processing time. Recently, a lot of research has been carried out in the DCT domain. DCT+Variance, DCT+Corr_Eng, DCT+EOL, and DCT+VOL are some prominent examples of DCT based methods.
Is an AI Coding Assistant Worth It in 2026?
Curious about the best AI coding assistant? An AI coding assistant is software that uses machine learning to help you get more done — it combines speed, accuracy, and an interface that just works. Hands-on testing shows real-world results vary, so a short free trial is the smartest way to decide. Whether you are a beginner or a pro, the right AI coding assistant slots into your workflow and pays for itself fast. This guide breaks down the top picks, their pros and cons, and who each one is best for.
Best AI Copywriting Tools in 2026
Looking for the best AI copywriting tool? An AI copywriting tool is software that uses machine learning to help you get more done — it can save you hours every week by automating repetitive work. Most options offer a generous free tier, with paid plans unlocking higher limits, faster processing, and team features. Whether you are a beginner or a pro, the right AI copywriting tool slots into your workflow and pays for itself fast. This guide breaks down the top picks, their pros and cons, and who each one is best for.
AI Paragraph Rewriters Reviews: What Actually Works in 2026
Looking for the best AI paragraph rewriter? An AI paragraph rewriter is software that uses machine learning to help you get more done — it can save you hours every week by automating repetitive work. Most options offer a generous free tier, with paid plans unlocking higher limits, faster processing, and team features. Whether you are a beginner or a pro, the right AI paragraph rewriter slots into your workflow and pays for itself fast. This guide breaks down the top picks, their pros and cons, and who each one is best for.
Isotropic position
In the fields of machine learning, the theory of computation, and random matrix theory, a probability distribution over vectors is said to be in isotropic position if its covariance matrix is proportional to the identity matrix. == Formal definitions == Let D {\textstyle D} be a distribution over vectors in the vector space R n {\textstyle \mathbb {R} ^{n}} . Then D {\textstyle D} is in isotropic position if, for vector v {\textstyle v} sampled from the distribution, E v v T = I d . {\displaystyle \mathbb {E} \,vv^{\mathsf {T}}=\mathrm {Id} .} A set of vectors is said to be in isotropic position if the uniform distribution over that set is in isotropic position. In particular, every orthonormal set of vectors is isotropic. As a related definition, a convex body K {\textstyle K} in R n {\textstyle \mathbb {R} ^{n}} is called isotropic if it has volume | K | = 1 {\textstyle |K|=1} , center of mass at the origin, and there is a constant α > 0 {\textstyle \alpha >0} such that ∫ K ⟨ x , y ⟩ 2 d x = α 2 | y | 2 , {\displaystyle \int _{K}\langle x,y\rangle ^{2}dx=\alpha ^{2}|y|^{2},} for all vectors y {\textstyle y} in R n {\textstyle \mathbb {R} ^{n}} ; here | ⋅ | {\textstyle |\cdot |} stands for the standard Euclidean norm.
Lise Getoor
Lise Getoor is an American computer scientist who is a distinguished professor and Baskin Endowed chair in the Computer Science and Engineering department, at the University of California, Santa Cruz, and an adjunct professor in the Computer Science Department at the University of Maryland, College Park. Her primary research interests are in machine learning and reasoning with uncertainty, applied to graphs and structured data. She also works in data integration, social network analysis and visual analytics. She has edited a book on Statistical relational learning that is a main reference in this domain. She has published many highly cited papers in academic journals and conference proceedings. She has also served as action editor for the Machine Learning Journal, JAIR associate editor, and TKDD associate editor. She received her Ph.D. from Stanford University, her M.S. from UC Berkeley, and her B.S. from UC Santa Barbara. Prior to joining University of California, Santa Cruz, she was a professor at the University of Maryland, College Park until November 2013. == Recognition == Getoor has multiple best paper awards, an NSF Career Award, and is an Association for the Advancement of Artificial Intelligence (AAAI) Fellow. In 2019, she was elected as an ACM Fellow "for contributions to machine learning, reasoning under uncertainty, and responsible data science", was selected as a Distinguished Alumna of the UC Santa Barbara Computer Science Department, was awarded the UCSC WiSE Chancellor's Achievement Award for Diversity, and was selected to give the UC Santa Cruz Faculty Research Lecture 2018-19, one of the highest recognitions given to UC faculty. She was named an IEEE Fellow in 2021, "for contributions to machine learning and reasoning under uncertainty". In October 2022, Getoor was elected a Fellow of the American Association for the Advancement of Science (AAAS). In 2024, she was named a Fellow of the American Academy of Arts and Sciences (AAA&S). Also in 2024, she received the ACM SIGKDD Innovation Award recognizing individuals with outstanding technical innovations in the field of Knowledge Discovery and Data Mining that have had a lasting impact in advancing the theory and practice of the field. == Personal life == Getoor's father was mathematician Ronald Getoor (1929–2017).