Develops path integral for spiked tensor model dynamics.
problem Dynamics of spiked tensor model with random initial conditions.
method Path integral approach applied to partial differential equations.
result Large-N saddle point equations dominated by melonic diagrams. FCNv2 robustness tested under noise and random initial conditions.
problem Assessing AI weather forecasting model robustness to input noise.
method Two experiments with varying noise levels and random initial conditions.
result FCNv2 preserves hurricane features under low to moderate noise, but underestimates intensity and persistence.
Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case o…
Study on the smoothness of solutions to a specific type of stochastic differential equation.
problem Regularity of solutions to mean-field G-SDEs. method Analysis of first and second order Fréchet differentiability in the random initial condition.
result Established the Fréchet differentiability of the solution and specified the corresponding equations.
In this paper we study a model of random knots obtained by fixing a space curve in n-dimensional Euclidean space with n>3, and orthogonally projecting the space curve on to random 3 dimensional subspaces. By varying the space curve we obtain different models of random parametrized knots, and we will study how the…
New method learns good initialization for gradient descent from past solutions.
problem Challenges in solving non-convex optimization problems.
method Learning good initialization rules from previous solutions.
result Our approach performs better than random initialization in various non-convex problems.
This work provides an additional step in the theoretical understanding of neural networks. We consider neural networks with one hidden layer and show that when learning symmetric functions, one can choose initial conditions so that standard SGD training efficiently produces generalization guarantees. We empirically ver…
Random Fourier features improve tabular deep learning convergence.
problem Tabular deep learning convergence issues.
method Random Fourier projections as a pre-processing step, projecting inputs into a fixed feature space.
result Random Fourier pre-processing accelerates tabular deep learning convergence.
Strong stability of ergodic iterations proven without ergodic driving sequence.
problem Ensuring strong stability of ergodic iterations under non-ergodic driving sequences.
method Revisiting processes driven by stationary ergodic sequences, proving strong stability under mild conditions on recursive maps.
result Strong stability of iterations proven without ergodic driving sequence.
Develops tests for conditional symmetry under group actions.
problem Testing conditional symmetry in distributions under group actions.
method Nonparametric randomization tests with kernel methods and asymptotic consistency.
result Tests achieve finite-sample Type I error control and power.
We investigate i.i.d. random complex dynamical systems generated by probability measures on finite unions of the loci of holomorphic families of rational maps on the Riemann sphere. We show that under certain conditions on the families, for a generic system, (especially, for a generic random polynomial dynamical system…
Gradient descent proves global convergence for 4-layer matrix factorization.
problem Global convergence of gradient descent on four-layer matrix factorization under random initialization.
method New techniques to show saddle-avoidance properties and extend eigenvalue theories.
result Polynomial-time global convergence guarantee for randomly initialized gradient descent on four-layer matrix factorization.
New algorithms use outsourced data to improve model training efficiency.
problem Limited computational resources restrict model training efficiency.
method Simulation-based algorithms using outsourced data to find good initial points.
result The algorithms can find good initial points with high probability under suitable conditions.
Paper refutes conjecture on tensor power iteration convergence in overcomplete models.
problem Understanding convergence of tensor power iteration in overcomplete random tensors.
method Analysis of tensor power iteration dynamics from random initialization.
result Polynomially many steps are necessary for convergence, refutes logarithmic conjecture.
The performance of the Self-Organizing Map (SOM) algorithm is dependent on the initial weights of the map. The different initialization methods can broadly be classified into random and data analysis based initialization approach. In this paper, the performance of random initialization (RI) approach is compared to that…
Study on neural network initialization with shaped infinite depth-and-width networks.
problem Understanding the distribution of random covariance matrices in shaped infinite-depth-and-width networks.
method Introduced the Neural Covariance SDE to model the distribution of the random covariance matrix.
result Identified the precise scaling of the activation function necessary for a non-trivial limit.
New method calculates geodesic distances in Gaussian random field manifolds.
problem Quantifying similarity between random fields in different regimes.
method Numerical method using geodesic distances in Gaussian random field manifolds.
result Estimation of geodesic distances for various initial conditions.
Gradient descent converges linearly for neural networks with specific conditions.
problem Optimizing neural networks with fixed width and depth.
method Local Polyak-Lojasiewicz criterion for gradient flow and descent.
result Gradient descent converges to zero-loss solutions under certain conditions.
GNNs with random node initialization are shown to be universally expressive.
problem Limitations of standard GNNs in distinguishing graphs.
method Random node initialization (RNI) to enhance GNNs' expressive power.
result GNNs with RNI are proven to be universally expressive.
New method shows random, diverse initializations are not essential for deep neural networks.
problem The necessity of random, diverse initializations in deep neural networks.
method Constructed a deep convolutional network with identical features by initializing weights to 0, enabling signal propagation and stable gradients.
result Random, diverse initializations are not necessary for training neural networks.
This paper examines the impact of random initialization in neural networks using NTK theory.
problem Understanding the impact of random initialization in neural networks using NTK theory.
method Analyzes the convergence of training dynamics and generalization error of wide neural networks with random initialization.
result The generalization error of wide neural networks trained by gradient descent is \( \Omega(n^{-\frac{3}{d+3}}) \), highlighting the benefits of mirror initialization and suggesting limitations of NTK theory.
Many neural networks use the tanh activation function, however when given a probability distribution as input, the problem of computing the output distribution in neural networks with tanh activation has not yet been addressed. One important example is the initialization of the echo state network in reservoir computing…
New method trains shallow neural networks with subquadratic width scaling.
problem Training shallow neural networks with optimal width scaling.
method Polyak-Lojasiewicz condition, smoothness, standard data assumptions, random matrix theory.
result Subquadratic scaling on network width with standard initialization strategies.
Randomly initialized transformers show extreme token preferences.
problem Structural biases in randomly initialized transformers.
method Dissection of transformer architecture at initialization.
result Initialization-induced biases persist throughout training.
The weights of a neural network are typically initialized at random, and one can think of the functions produced by such a network as having been generated by a prior over some function space. Studying random networks, then, is useful for a Bayesian understanding of the network evolution in early stages of training. In…
Bounds neural network output distribution to Gaussian for random initialization.
problem Quantifying the distribution of randomly initialized deep neural networks.
method Quantitative Gaussian approximation using quadratic Wasserstein distance.
result Explicit inequalities show how network sizes affect Gaussian behavior.
We study exponential Levy models with change-point which is a random variable, independent from initial Levy processes. On canonical space with initially enlarged filtration we describe all equivalent martingale measures for change-point model and we give the conditions for the existence of f-divergence minimal equival…
Gradient Descent with small random initialization solves rank-1 matrix completion efficiently.
problem Matrix completion for rank-1 symmetric matrices.
method Gradient Descent with small random initialization.
result Gradient Descent converges to the ground truth for rank-1 symmetric matrix completion.
Study uses random matrix theory to improve tensor approximation accuracy.
problem Improving tensor approximation accuracy in the presence of noise.
method Random matrix theory applied to tensor unfoldings.
result Characterizes spectral behavior of tensor unfoldings and predicts reconstruction performance.
RSIC identifies multiple ranks of interest in NMF by analyzing residual sensitivity.
problem Determining the optimal rank in NMF.
method RSIC analyzes sensitivity of relative residuals to different initializations.
result RSIC identifies meaningful ranks consistent with data structure.
Batch normalization prevents rank collapse in deep networks, improving training stability.
problem Rank collapse in randomly initialized deep networks with increasing depth.
method Investigates spectral instabilities in random matrices and uses batch normalization to avoid rank collapse.
result Batch normalization prevents rank collapse in both linear and ReLU networks, improving training stability.
This paper improves privacy bounds for DP algorithms using f-DP.
problem Difficulty in analyzing randomness in DP algorithms due to mixture distributions.
method Derives a closed-form expression for trade-off functions and analyzes f-DP. result Enhances privacy of DP-GD with random initialization and shuffling models.
Two new scalable K-means initialization methods proposed for large-scale clustering.
problem Efficient initialization for large-scale clustering problems.
method Divide-and-conquer approach and random projection method for multiple lower-dimensional subspaces.
result The proposed methods outperform state-of-the-art in large-scale clustering tasks.
We analyze the disordered Riemannian geometry resulting from random perturbations of the Euclidean metric. We focus on geodesics, the paths traced out by a particle traveling in this quenched random environment. By taking the point of the view of the particle, we show that the law of its observed environment is absolut…
Recent pruning methods at initialization fall short of random pruning's accuracy.
problem Improving neural network accuracy through pruning at initialization.
method Various pruning methods (SNIP, GraSP, SynFlow, magnitude pruning) are evaluated; per-layer pruning decisions are proposed.
result Randomly shuffling or sampling initial weights preserves or improves accuracy, suggesting challenges with pruning heuristics.
Paper proposes a new method for supervised manifold learning using random forest proximities.
problem Existing supervised manifold learning methods fail to uncover meaningful embeddings due to using class-conditional distances.
method Proposes a data-geometry-preserving variant of random forest proximities as an initialization for manifold learning methods.
result Local and global structure preservation is near universal across manifold learning approaches using diffusion-based algorithms.
Recursive stochastic algorithms have gained significant attention in the recent past due to data driven applications. Examples include stochastic gradient descent for solving large-scale optimization problems and empirical dynamic programming algorithms for solving Markov decision problems. These recursive stochastic a…
We study the strong predictable representation property in filtrations initially enlarged with a random variable L. We prove that the strong predictable representation property can always be transferred to the enlarged filtration as long as the classical density hypothesis of Jacod (1985) holds. This generalizes the ex…
The study analyzes deep linear networks from random initialization, capturing dynamics and hyperparameter effects.
problem Understanding training dynamics in deep linear networks from random initialization.
method Theoretical analysis of gradient descent dynamics in deep linear networks with random initialization and large data.
result Captures the 'wider is better' effect and hyperparameter transfer effects, contrasting with neural-tangent parameterization.
Study on adversarial robustness in neural networks across initialization and training phases.
problem Understanding adversarial robustness in neural networks during different learning stages.
method Analyzes adversarial robustness in various scenarios of over-parameterized networks with quadratic targets and infinite samples.
result Robustness can worsen when test error improves, and vice versa, revealing new tradeoffs.
Data-driven discovery of differential equations has been an emerging research topic. We propose a novel algorithm subsampling-based threshold sparse Bayesian regression (SubTSBR) to tackle high noise and outliers. The subsampling technique is used for improving the accuracy of the Bayesian learning algorithm. It has tw…
Randomly initialized neural networks can linearly separate arbitrary sets.
problem Mapping two arbitrary sets to linearly separable sets.
method Randomly initialized one-layer neural networks with sufficient width.
result With high probability, these networks can transform two sets into linearly separable sets.
AMP method reconstructs rank-one matrices from noisy data efficiently.
problem Reconstructing rank-one matrices with prior structural information from noisy observations.
method Approximate Message Passing (AMP) with random initialization.
result AMP from random initialization converges rapidly and globally.
KSS method converges and recovers correct clustering under certain conditions.
problem Subspace clustering for semi-randomly sampled data.
method Local convergence analysis and recovery guarantee for KSS method.
result KSS method converges superlinearly and finds correct clustering within loglog N iterations.
Paper presents a new backward deep BSDE method for solving nonlinear FBSDE problems.
problem Nonlinear Forward Backward Stochastic Differential Equations (FBSDE) with terminal conditions.
method Backward deep BSDE method applied to FBSDE with nonlinear generators and random initial conditions.
result Derives exact and Taylor-based approximations for time-stepping nonlinear BSDEs.
Algorithm recovers factors of rank-1 matrices from noisy measurements.
problem Estimating factors of a rank-1 matrix from nonlinearly transformed and noisy measurements.
method Alternating minimization with random initialization and analysis of empirical error recursion.
result Algorithm converges geometrically fast from random initialization, with sharp guarantees.
Study shows different initialization schemes for LoRA finetuning impact performance.
problem The impact of initialization schemes on LoRA finetuning performance.
method Compared two initialization schemes: B=0, A=random vs. A=0, B=random.
result First initialization scheme yields better performance on average.
The Residual Network (ResNet), proposed in He et al. (2015), utilized shortcut connections to significantly reduce the difficulty of training, which resulted in great performance boosts in terms of both training and generalization error. It was empirically observed in He et al. (2015) that stacking more layers of resid…