Neural networks memorize exceptions, leading to poor generalization.
problem Memorization of exceptions hinders neural network generalization.
method Formalized memorization-generalization interplay, proposed MAT to shift logits.
result MAT improves generalization by learning robust patterns invariant across distributions.
Local data coverage governs memorization in diffusion models.
problem Memorization in diffusion models
method Derive a theoretical criterion based on local data coverage
result Predicts memorization based on density of training data in neighborhood and dataset size
Geometric framework explains memorization in generative models.
problem Memorization in generative models raises legal and privacy concerns.
method Manifold memorization hypothesis (MMH) using manifold geometry.
result Formalizes and categorizes memorization into overfitting and distribution-driven types.
The paper identifies patterns in language model weights used for memorizing paragraphs.
problem Locating the specific mechanisms and weights used by language models to memorize paragraphs.
method Examined gradients and attention patterns in language models to identify memorized paragraphs.
result Gradients of memorized paragraphs have a distinguishable spatial pattern, and localized attention heads are involved in paragraph memorization.
Consistency distillation reduces memorization in diffusion models without harming sample quality.
problem Understanding how distillation affects memorization in diffusion models.
method Analysis of consistency distillation in diffusion models using a random feature neural network model.
result Consistency distillation reduces memorization in diffusion models without harming sample quality.
Study shows how deep generative models can memorize data.
problem Understanding and preventing memorization in deep generative models.
method Adapted a memorization measure for unsupervised density estimation and demonstrated its effectiveness.
result Memorization in deep generative models differs from mode collapse and overfitting.
We study finite sample expressivity, i.e., memorization power of ReLU networks. Recent results require N N N hidden nodes to memorize/interpolate arbitrary N N N data points. In contrast, by exploiting depth, we show that 3-layer ReLU networks with Ω ( N ) Ω(\sqrt{N}) Ω ( N ) hidden nodes can perfectly memorize most datasets with N N N po…
Memorizing rare examples helps neural networks generalize better.
problem Improving generalization in deep learning models.
method Theoretical analysis and experiments on neural networks with composition capability.
result Memorizing rare examples can help neural networks make correct predictions on rare test examples.
Optimal ReLU networks can memorize any separable set of points with a small number of parameters.
problem The optimal number of parameters required to memorize a set of points using ReLU networks.
method Construction of ReLU networks with specific bit complexity to memorize points satisfying a mild separability assumption.
result Optimal ReLU networks can memorize any separable set of points with a number of parameters that is i l d e O ( N ) ilde{O}(\sqrt{N}) i l d e O ( N ) . LLMs can memorize economic data and recall exact values before their training cutoff.
problem Evaluating the trustworthiness of LLMs' economic forecasts during their training period.
method Demonstrated through counterfactual forecasting and analysis of LLMs' recall ability.
result LLMs have memorized economic and financial data, leading to recall-level accuracy before their knowledge cutoff.
New method reduces memorization in diffusion models without sacrificing image quality.
problem Diffusion models often memorize training data, especially with small datasets.
method Train models using noisy data at large noise scales to reduce memorization.
result Significant reduction in memorization without compromising image quality.
Logarithmic network width suffices for robust memorization.
problem Achieving robust memorization in neural networks.
method Established upper and lower bounds on robust memorization radius.
result Width logarithmic in the number of samples is necessary and sufficient for robust memorization.
New approach shows data memorization trade-offs in large models.
problem Data memorization in large language models and its privacy implications.
method Developed a new approach using strong data processing inequalities to prove lower bounds on memorization.
result Proved that Ω ( d ) Ω(d) Ω ( d ) bits of training data information must be memorized for O ( 1 ) O(1) O ( 1 ) examples, decaying with example growth. Study compares memorization of SimCLR to supervised and random labels training.
problem Understanding memorization in contrastive learning.
method Investigated SimCLR's memorization properties compared to supervised and random labels training.
result SimCLR's memorization is similar to random labels training in terms of training object complexity distribution.
Generative diffusion models gradually memorize training data, losing independent dimensions.
problem Understanding how generative diffusion models memorize training data, especially on low-dimensional manifolds.
method Measuring latent dimensionality via the learned score field, proposing a geometric memorization theory.
result Generative diffusion models experience a smooth collapse of their capacity to vary across independent directions as data become scarce, leading to near point-wise replication of salient features.
Deep networks preferentially learn shared features, avoiding memorization in early layers.
problem Understanding how deep neural networks generalize vs. memorize training data.
method Replica-based mean field geometric analysis of deep neural networks.
result Deep layers predominantly memorize, while early layers are minimally affected.
Deep networks can memorize random labels; symmetric loss improves this.
problem Deep networks can memorize random labels, ignoring standard regularization.
method Empirical studies with MNIST and CIFAR-10 datasets, formal definition of robustness.
result Symmetric loss function improves network's ability to resist memorization.
This paper explores memorization in adversarial training and proposes a mitigation algorithm.
problem Understanding and mitigating robust overfitting in adversarial training.
method Demonstrated the capacity of deep networks to memorize adversarial examples, analyzed convergence and generalization issues, and proposed a new mitigation algorithm.
result Identified robust overfitting as a significant drawback of adversarial training and proposed a mitigation algorithm.
Study shows FL reduces unintended memorization by clustering data and using strong user-level privacy.
problem Unintended memorization in federated learning.
method Examined the effect of clustering data and using strong user-level differential privacy in FL.
result Clustering data and strong user-level differential privacy reduce unintended memorization.
ResMem improves model generalization by explicitly memorizing residuals.
problem Improving model generalization in neural networks.
method ResMem algorithm that augments a model with a k-nearest neighbor based regressor fitted to residuals.
result ResMem consistently improves test set generalization across various benchmarks.
Diffusion models can memorize training data, limiting their creativity and privacy.
problem Memorization in diffusion models that reproduces training data instead of generating novel outputs.
method Dual-separation approach via statistical estimation and network approximation.
result Pruning-based method reduces memorization while maintaining generation quality.
RAF model explains neural networks' dual rule learning and fact memorization.
problem Understanding how neural networks learn rules and memorize facts simultaneously.
method Introduces the Rules-and-Facts (RAF) model to bridge generalization and memorization.
result Characterizes conditions for simultaneous rule learning and fact memorization in neural networks.
Learning to solve sequential tasks with recurrent models requires the ability to memorize long sequences and to extract task-relevant features from them. In this paper, we study the memorization subtask from the point of view of the design and training of recurrent neural networks. We propose a new model, the Linear Me…
Gradient descent memorizes many Gaussians efficiently.
problem Memorizing many Gaussians with minimal parameters.
method Gradient descent on a depth-two neural network.
result One step of gradient descent memorizes $Ω\left(\frac{dq}{\log^4(d)}
ight)$ Gaussians.
Introduces Relational Privacy (RP) to control relation memorization in question answering models.
problem Relation memorization in question answering models can lead to privacy issues.
method Formalizes Relational Privacy (RP) and Differential Relational Privacy (DrP), providing bounds on relation memorization.
result DrP allows effective learning of general properties of underlying concepts while preventing relation memorization.
The bias potential model explains how generative models can generalize or memorize samples.
problem Understanding and achieving generalization in generative models like GANs.
method Introducing the bias potential model to analyze the behavior of generative models.
result Dimension-independent generalization accuracy can be achieved with early stopping in the bias potential model.
The study uncovers the conditions under which diffusion models memorize or generalize.
problem Understanding the balance between memorization and generalization in diffusion models.
method Theoretical and mathematical framework to investigate memorization and generalization in diffusion models.
result Theoretical crossover point predicts a phase transition in diffusion models, validating the hypothesis.
State-of-the-art results on image recognition tasks are achieved using over-parameterized learning algorithms that (nearly) perfectly fit the training set and are known to fit well even random labels. This tendency to memorize the labels of the training data is not explained by existing theoretical analyses. Memorizati…
Deep models memorize training data in geophysical inversion, leading to biased posterior distributions.
problem Memorization of training data biases learned priors in geophysical inverse problems.
method Casting generative models' training as maximum likelihood, we show memorization results in a reweighted empirical distribution for diffusion models, leading to Gaussian mixture priors and posteriors.
result Memorization leads to posterior distributions that are likelihood-weighted lookup among stored training examples, affecting full waveform inversion outcomes.
New training method for ReLU networks achieves optimal weight size for memorization.
problem Approximate memorization of arbitrary real labels with neural networks.
method Complex recombination training procedure for ReLU networks.
result Approximate memorization with nearly optimal weight size and neuron count.
Diffusion models generalize well until a threshold is reached, preventing memorization.
problem Understanding why diffusion models don't memorize training data.
method Investigation of training dynamics and two timescales: τ g e n τ_\mathrm{gen} τ gen and τ m e m τ_\mathrm{mem} τ mem . result The threshold τ m e m τ_\mathrm{mem} τ mem increases linearly with training set size n n n , preventing memorization. Large learning rates prevent memorization in denoising score matching.
problem Memorization of training data in diffusion-based generative models.
method Investigating the role of large learning rates in the small-noise regime, proving that they prevent convergence to the empirical optimal score.
result Large learning rates prevent memorization by making it impossible for the learned score to be arbitrarily close to the empirical optimal score.
Transformers with CoT don't enhance reasoning power across all tasks.
problem Does CoT enhance the reasoning power of transformers?
method Examined the memorization capabilities of fixed-precision transformers with and without CoT.
result Transformers with CoT cannot memorize all reasoning tasks, leading to a negative answer.
Study interprets neural network generalization and memorization on corrupted data.
problem Understanding when a neural network has memorized corrupted data versus learned the underlying rule.
method Analyzes multi-layer perceptrons and Transformers on modular arithmetic tasks with corrupted labels.
result Regularization methods can force networks to ignore corrupted data, improving accuracy on uncorrupted data.
Early stopping improves generalization in overparameterized diffusion models.
problem Understanding and optimizing generalization in overparameterized diffusion models.
method Revisiting diffusion models, showing generalization occurs before memorization, and developing a phase diagram.
result Generalization time scales with dataset size, supporting early-stopping criteria.
New experiments show deep networks benefit from memorizing rare data points.
problem Understanding why neural networks perform well on training data.
method Estimating influence and memorization values of training examples.
result Deep networks benefit from memorizing rare data points for good performance.
The roles played by learning and memorization represent an important topic in deep learning research. Recent work on this subject has shown that the optimization behavior of DNNs trained on shuffled labels is qualitatively different from DNNs trained with real labels. Here, we propose a novel permutation approach that …
We improve deep threshold networks' memorization capacity exponentially.
problem Memorizing datasets with randomized labels using deep neural networks.
method Using Gaussian random weights in the first layer and binary or integer weights in subsequent layers, we prove a new dependence on minimum distance.
result We show that O ~ ( 1 δ + n ) \widetilde{\mathcal{O}}(\frac{1}{\delta} + \sqrt{n}) O ( δ 1 + n ) neurons and O ~ ( d δ + n ) \widetilde{\mathcal{O}}(\frac{d}{\delta} + n) O ( δ d + n ) weights are sufficient. Given data with noisy labels, over-parameterized deep networks can gradually memorize the data, and fit everything in the end. Although equipped with corrections for noisy labels, many learning methods in this area still suffer overfitting due to undesired memorization. In this paper, to relieve this issue, we propose …
The IM effect helps detect anomalies by memorizing inliers early.
problem Challenges in fully unsupervised outlier detection.
method Theoretical study of a simple autoencoder, focusing on early training dynamics.
result Characterization of IM effect emergence, strength, and persistence.
Over-parameterized models can memorize noisy labels and still generalize well, revealing a hidden structure.
problem Understanding how over-parameterized models can simultaneously memorize noisy labels and generalize well.
method Investigated through modular arithmetic tasks with label noise using two-layer neural networks.
result Over-parameterized models can achieve near-perfect test accuracy with 80% label noise by extracting an internal generalization structure.
Data selection boosts fact memorization in language models.
problem Language models struggle to accurately memorize factual knowledge.
method Formalizes fact memorization, proposes data selection schemes based on training loss.
result Data selection boosts fact accuracy to model capacity and improves performance.
New NTK bounds show deep networks with minimum over-parameterization can still memorize and optimize.
problem Understanding memorization and optimization in sub-linear over-parameterized deep networks.
method Lower bound on NTK eigenvalues for deep networks with minimum over-parameterization.
result Deep networks with minimum over-parameterization can still be powerful memorizers and optimizers.
Changing initialization scale affects deep model generalization, leading to memorization or improved performance.
problem Understanding how initialization scale impacts deep model generalization and memorization.
method Experimental setup with varying initialization scales, analysis of activation and loss functions, and development of an alignment measure.
result Increasing initialization scale leads to memorization, and decreasing it improves generalization, depending on activation and loss functions.
We propose a new notion of `non-linearity' of a network layer with respect to an input batch that is based on its proximity to a linear system, which is reflected in the non-negative rank of the activation matrix. We measure this non-linearity by applying non-negative factorization to the activation matrix. Considering…
The study finds that memorization is necessary or harmful depending on the prior distribution and noise level.
problem The impact of memorization on generalization in overparameterized models.
method An overparameterized linear model with general priors in a Bayesian setup.
result Explicit conditions for optimal generalization based on the prior distribution and noise level.
The paper analyzes how deep models memorize spurious features.
problem Understanding how deep models memorize spurious features in training data.
method Characterizes spurious feature memorization via model stability and feature alignment.
result Memorization of spurious features weakens as generalization capability increases.
Deep neural networks (DNNs) can easily fit a random labeling of the training data with zero training error. What is the difference between DNNs trained with random labels and the ones trained with true labels? Our paper answers this question with two contributions. First, we study the memorization properties of DNNs. O…