arXiv subjects

Mihai Nica

Publications and source records attributed to Mihai Nica.

At least 19 recordsLinked to original sources

arXiv2026

Geometric Dyson Brownian Motions and the Free Log-Normal Limit for a Non-Square Gaussian Matrix Product

We study the squared singular value spectrum of a non-square product of independent real Gaussian matrices, equivalently the feature covariance spectrum of a deep linear neural network at initialization. Starting from the fixed-$m$ covariance diffusion previously obtained in the proportional depth-width limit, we record an equivalent matrix realization, describe its affine invariance, and derive the interacting diffusion satisfied by its eigenvalues. We then take a second limit, sending $m\to\infty$ on the accelerated spectral clock $τ=mt$, which corresponds in this sequential construction to the relation $dm/n\to\barτ$. We establish convergence of the empirical spectral measure path to a deterministic mean-field limit and derive a closed Burgers equation for its $T$-transform. Together with the proportional depth-width limit, these results give a rigorous sequential route from the deep non-square Gaussian product to the free log-normal limit of its feature covariance spectrum; for more general initial laws, the transform yields a free multiplicative convolution form. We further analyze the support of the free log-normal law, give a fixed point iteration for numerical evaluation and a formal Marchenko--Pastur approximation at small time, and use the limiting spectrum to predict the risk in a toy random feature model.

math.PR↗

arXiv2025

Optimal Strategy in "Guess Who?": Beyond Binary Search

"Guess Who?" is a popular two player game where players ask "Yes"/"No" questions to search for their opponent's secret identity from a pool of possible candidates. This is modeled as a simple stochastic game. Using this model, the optimal strategy is explicitly found. Contrary to popular belief, performing a binary search is \emph{not} always optimal. Instead, the optimal strategy for the player who trails is to make certain bold plays in an attempt catch up. This is discovered by first analyzing a continuous version of the game where players play indefinitely and the winner is never decided after finitely many rounds.

math.PR↗

arXiv2025

A Gaussian integral formula for the Hermite polynomials: Combinatorics, Asymptotics and Applications

The Hermite polynomials are ubiquitous but can be difficult to work with due to their unwieldy definition in terms of derivatives. To remedy this, we showcase an underappreciated Gaussian integral formula for the Hermite polynomials, which is especially useful for generalizing to multivariable Hermite polynomials. Taking this as our definition, we prove many useful consequences, including: 1. Combinatorial interpretations for the Hermite polynomials, including a proof of orthogonality. 2. A more elementary proof of Plancherel--Rotach asymptotics that does not involve residues. 3. Limit theorems for GUE random matrices and Dyson's Brownian motion, including bulk convergence to the semi-circle law and edge convergence to the Airy limit/Tracy-Widom law. 4. An analysis of a phase transition in the spiked GUE random matrix as the top eigenvalue goes from well-separated to attached to the bulk, analogous to the BBP phase transition. 5. Elementary derivations of Edgeworth expansions and multivariable Edgeworth expansions. This article is primarily expository and features many illustrative figures.

math.PR↗

arXiv2025

The generalized Alice HH vs Bob HT problem

In 2024, Daniel Litt posed a simple coinflip game pitting Alice's "Heads-Heads" vs Bob's "Heads-Tails": who is more likely to win if they score 1 point per occurrence of their substring in a sequence of n fair coinflips? This attracted over 1 million views on X and quickly spawned several articles explaining the counterintuitive solution. We study the generalized game, where the set of coin outcomes, {Heads, Tails}, is generalized to an arbitrary finite alphabet A, and where Alice's and Bob's substrings are any finite A-strings of the same length. We find that the winner of Litt's game can be determined by a single quantity which measures the amount of prefix/suffix self-overlaps in each string; whoever's string has more overlaps loses. For example, "Heads-Tails" beats "Heads-Heads" in the original problem because "Heads-Heads" has a prefix/suffix overlap of length 1 while "Heads-Tails" has none. The method of proof is to develop a precise Edgeworth expansion for discreteMarkov chains, and apply this to calculate Alice's and Bob's probability to win the game correct to order O(1/n).

math.PR↗

arXiv2024

Depth Degeneracy in Neural Networks: Vanishing Angles in Fully Connected ReLU Networks on Initialization

Despite remarkable performance on a variety of tasks, many properties of deep neural networks are not yet theoretically understood. One such mystery is the depth degeneracy phenomenon: the deeper you make your network, the closer your network is to a constant function on initialization. In this paper, we examine the evolution of the angle between two inputs to a ReLU neural network as a function of the number of layers. By using combinatorial expansions, we find precise formulas for how fast this angle goes to zero as depth increases. These formulas capture microscopic fluctuations that are not visible in the popular framework of infinite width limits, and leads to qualitatively different predictions. We validate our theoretical results with Monte Carlo experiments and show that our results accurately approximate finite network behaviour. \review{We also empirically investigate how the depth degeneracy phenomenon can negatively impact training of real networks.} The formulas are given in terms of the mixed moments of correlated Gaussians passed through the ReLU function. We also find a surprising combinatorial connection between these mixed moments and the Bessel numbers that allows us to explicitly evaluate these moments.

stat.ML↗

arXiv2024

Differential Equation Scaling Limits of Shaped and Unshaped Neural Networks

Recent analyses of neural networks with shaped activations (i.e. the activation function is scaled as the network size grows) have led to scaling limits described by differential equations. However, these results do not a priori tell us anything about "ordinary" unshaped networks, where the activation is unchanged as the network size grows. In this article, we find similar differential equation based asymptotic characterization for two types of unshaped networks. Firstly, we show that the following two architectures converge to the same infinite-depth-and-width limit at initialization: (i) a fully connected ResNet with a $d^{-1/2}$ factor on the residual branch, where $d$ is the network depth. (ii) a multilayer perceptron (MLP) with depth $d \ll$ width $n$ and shaped ReLU activation at rate $d^{-1/2}$. Secondly, for an unshaped MLP at initialization, we derive the first order asymptotic correction to the layerwise correlation. In particular, if $ρ_\ell$ is the correlation at layer $\ell$, then $q_t = \ell^2 (1 - ρ_\ell)$ with $t = \frac{\ell}{n}$ converges to an SDE with a singularity at $t=0$. These results together provide a connection between shaped and unshaped network architectures, and opens up the possibility of studying the effect of normalization methods and how it connects with shaping activation functions.

stat.ML↗

arXiv2024

Dynamic Sparse Training with Structured Sparsity

Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically less computationally expensive, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity by imposing a constant fan-in constraint. Using our empirical analysis of existing DST methods at high sparsity, we additionally employ a neuron ablation method which enables SRigL to achieve state-of-the-art sparse-to-sparse structured DST performance on a variety of Neural Network (NN) architectures. Using a 90% sparse linear layer, we demonstrate a real-world acceleration of 3.4x/2.5x on CPU for online inference and 1.7x/13.0x on GPU for inference with a batch size of 256 when compared to equivalent dense/unstructured (CSR) sparse layers, respectively.

cs.LG↗

arXiv2023

Population mobility, well-mixed clustering and disease spread: a look at COVID-19 Spread in the United States and preventive policy insights

The epidemiology of pandemics is classically viewed using geographical and political borders; however, these artificial divisions can result in a misunderstanding of the current epidemiological state within a given region. To improve upon current methods, we propose a clustering algorithm which is capable of recasting regions into well-mixed clusters such that they have a high level of interconnection while minimizing the external flow of the population towards other clusters. Moreover, we analyze and identify so called core clusters, clusters that retain their features over time (temporally stable) and independent of the presence or absence of policy measures. In order to demonstrate the capabilities of this algorithm, we use US county-level cellular mobility data to divide the country into such clusters. Herein, we show a more granular spread of SARS-CoV-2 throughout the first weeks of the pandemic. Moreover, we are able to identify areas (groups of counties) that were experiencing above average levels of transmission within a state, as well as pan-state areas (clusters overlapping more than one state) with very similar disease spread. Therefore, our method enables policymakers to make more informed decisions on the use of public health interventions within their jurisdiction, as well as guide collaboration with surrounding regions to benefit the general population in controlling the spread of communicable diseases.

q-bio.PE↗

arXiv2023

Bandit-Driven Batch Selection for Robust Learning under Label Noise

We introduce a novel approach for batch selection in Stochastic Gradient Descent (SGD) training, leveraging combinatorial bandit algorithms. Our methodology focuses on optimizing the learning process in the presence of label noise, a prevalent issue in real-world datasets. Experimental evaluations on the CIFAR-10 dataset reveal that our approach consistently outperforms existing methods across various levels of label corruption. Importantly, we achieve this superior performance without incurring the computational overhead commonly associated with auxiliary neural network models. This work presents a balanced trade-off between computational efficiency and model efficacy, offering a scalable solution for complex machine learning applications.

cs.LG↗

arXiv2023

Diffusion on the Probability Simplex

Diffusion models learn to reverse the progressive noising of a data distribution to create a generative model. However, the desired continuous nature of the noising process can be at odds with discrete data. To deal with this tension between continuous and discrete objects, we propose a method of performing diffusion on the probability simplex. Using the probability simplex naturally creates an interpretation where points correspond to categorical probability distributions. Our method uses the softmax function applied to an Ornstein-Unlenbeck Process, a well-known stochastic differential equation. We find that our methodology also naturally extends to include diffusion on the unit cube which has applications for bounded image generation.

cs.LG↗

arXiv2023

Using alternating de Bruijn sequences to construct de Bruijn tori

A de Bruijn torus is the two dimensional generalization of a de Bruijn sequence. While some methods exist to generate these tori, only a few methods of construction are known. We present a novel method to generate de Bruijn tori with rectangular windows by combining two variants de Bruijn sequences called `Alternating de Bruijn sequences' and `De Bruijn families'.

math.CO↗

arXiv2023

The Neural Covariance SDE: Shaped Infinite Depth-and-Width Networks at Initialization

The logit outputs of a feedforward neural network at initialization are conditionally Gaussian, given a random covariance matrix defined by the penultimate layer. In this work, we study the distribution of this random matrix. Recent work has shown that shaping the activation function as network depth grows large is necessary for this covariance matrix to be non-degenerate. However, the current infinite-width-style understanding of this shaping method is unsatisfactory for large depth: infinite-width analyses ignore the microscopic fluctuations from layer to layer, but these fluctuations accumulate over many layers. To overcome this shortcoming, we study the random covariance matrix in the shaped infinite-depth-and-width limit. We identify the precise scaling of the activation function necessary to arrive at a non-trivial limit, and show that the random covariance matrix is governed by a stochastic differential equation (SDE) that we call the Neural Covariance SDE. Using simulations, we show that the SDE closely matches the distribution of the random covariance matrix of finite networks. Additionally, we recover an if-and-only-if condition for exploding and vanishing norms of large shaped networks based on the activation function.

stat.ML↗

arXiv2023

Network Degeneracy as an Indicator of Training Performance: Comparing Finite and Infinite Width Angle Predictions

Neural networks are powerful functions with widespread use, but the theoretical behaviour of these functions is not fully understood. Creating deep neural networks by stacking many layers has achieved exceptional performance in many applications and contributed to the recent explosion of these methods. Previous works have shown that depth can exponentially increase the expressibility of the network. However, as networks get deeper and deeper, they are more susceptible to becoming degenerate. We observe this degeneracy in the sense that on initialization, inputs tend to become more and more correlated as they travel through the layers of the network. If a network has too many layers, it tends to approximate a (random) constant function, making it effectively incapable of distinguishing between inputs. This seems to affect the training of the network and cause it to perform poorly, as we empirically investigate in this paper. We use a simple algorithm that can accurately predict the level of degeneracy for any given fully connected ReLU network architecture, and demonstrate how the predicted degeneracy relates to training dynamics of the network. We also compare this prediction to predictions derived using infinite width networks.

cs.LG↗

arXiv2023

Uniform convergence to the Airy line ensemble

We show that classical integrable models of last passage percolation and the related nonintersecting random walks converge uniformly on compact sets to the Airy line ensemble. Our core approach is to show convergence of nonintersecting Bernoulli random walks in all feasible directions in the parameter space. We then use coupling arguments to extend convergence to other models.

math.PR↗

arXiv2022

Bounding generalization error with input compression: An empirical study with infinite-width networks

Estimating the Generalization Error (GE) of Deep Neural Networks (DNNs) is an important task that often relies on availability of held-out data. The ability to better predict GE based on a single training set may yield overarching DNN design principles to reduce a reliance on trial-and-error, along with other performance assessment advantages. In search of a quantity relevant to GE, we investigate the Mutual Information (MI) between the input and final layer representations, using the infinite-width DNN limit to bound MI. An existing input compression-based GE bound is used to link MI and GE. To the best of our knowledge, this represents the first empirical study of this bound. In our attempt to empirically falsify the theoretical bound, we find that it is often tight for best-performing models. Furthermore, it detects randomization of training labels in many cases, reflects test-time perturbation robustness, and works well given only few training samples. These results are promising given that input compression is broadly applicable where MI can be estimated with confidence.

cs.LG↗

arXiv2022

The Exponentially Tilted Gaussian Prior for Variational Autoencoders

An important property for deep neural networks is the ability to perform robust out-of-distribution detection on previously unseen data. This property is essential for safety purposes when deploying models for real world applications. Recent studies show that probabilistic generative models can perform poorly on this task, which is surprising given that they seek to estimate the likelihood of training data. To alleviate this issue, we propose the exponentially tilted Gaussian prior distribution for the Variational Autoencoder (VAE) which pulls points onto the surface of a hyper-sphere in latent space. This achieves state-of-the art results on the area under the curve-receiver operator characteristics metric using just the log-likelihood that the VAE naturally assigns. Because this prior is a simple modification of the traditional VAE prior, it is faster and easier to implement than competitive methods.

cs.LG↗

arXiv2021

The Future is Log-Gaussian: ResNets and Their Infinite-Depth-and-Width Limit at Initialization

Theoretical results show that neural networks can be approximated by Gaussian processes in the infinite-width limit. However, for fully connected networks, it has been previously shown that for any fixed network width, $n$, the Gaussian approximation gets worse as the network depth, $d$, increases. Given that modern networks are deep, this raises the question of how well modern architectures, like ResNets, are captured by the infinite-width limit. To provide a better approximation, we study ReLU ResNets in the infinite-depth-and-width limit, where both depth and width tend to infinity as their ratio, $d/n$, remains constant. In contrast to the Gaussian infinite-width limit, we show theoretically that the network exhibits log-Gaussian behaviour at initialization in the infinite-depth-and-width limit, with parameters depending on the ratio $d/n$. Using Monte Carlo simulations, we demonstrate that even basic properties of standard ResNet architectures are poorly captured by the Gaussian limit, but remarkably well captured by our log-Gaussian limit. Moreover, our analysis reveals that ReLU ResNets at initialization are hypoactivated: fewer than half of the ReLUs are activated. Additionally, we calculate the interlayer correlations, which have the effect of exponentially increasing the variance of the network output. Based on our analysis, we introduce Balanced ResNets, a simple architecture modification, which eliminates hypoactivation and interlayer correlations and is more amenable to theoretical analysis.

stat.ML↗

arXiv2021

Finding Critical Scenarios for Automated Driving Systems: A Systematic Literature Review

Scenario-based approaches have been receiving a huge amount of attention in research and engineering of automated driving systems. Due to the complexity and uncertainty of the driving environment, and the complexity of the driving task itself, the number of possible driving scenarios that an ADS or ADAS may encounter is virtually infinite. Therefore it is essential to be able to reason about the identification of scenarios and in particular critical ones that may impose unacceptable risk if not considered. Critical scenarios are particularly important to support design, verification and validation efforts, and as a basis for a safety case. In this paper, we present the results of a systematic literature review in the context of autonomous driving. The main contributions are: (i) introducing a comprehensive taxonomy for critical scenario identification methods; (ii) giving an overview of the state-of-the-art research based on the taxonomy encompassing 86 papers between 2017 and 2020; and (iii) identifying open issues and directions for further research. The provided taxonomy comprises three main perspectives encompassing the problem definition (the why), the solution (the methods to derive scenarios), and the assessment of the established scenarios. In addition, we discuss open research issues considering the perspectives of coverage, practicability, and scenario space explosion.

cs.SE↗