Search Results for author:
Found 39 papers, 17 papers with code
Localizing Paragraph Memorization in Language Models
Can we localize the weights and mechanisms used by a language model to memorize and recite entire paragraphs of its training data?
How Private is DP-SGD?
We demonstrate a substantial gap between the privacy guarantees of the Adaptive Batch Linear Queries (ABLQ) mechanism under different types of batch sampling: (i) Shuffling, and (ii) Poisson subsampling; the typical analysis of Differentially Private Stochastic Gradient Descent (DP-SGD) follows by interpreting it as a post-processing of ABLQ.
Training Differentially Private Ad Prediction Models with Semi-Sensitive Features
Motivated by problems arising in digital advertising, we introduce the task of training differentially private (DP) machine learning models with semi-sensitive features.
Optimal Unbiased Randomizers for Regression with Label Differential Privacy
We propose a new family of label randomizers for training regression models under the constraint of label differential privacy (DP).
Can Neural Network Memorization Be Localized?
Recent efforts at explaining the interplay of memorization and generalization in deep overparametrized networks have posited that neural networks $\textit{memorize}$ "hard" examples in the final few layers of the model.
Ticketed Learning-Unlearning Schemes
Subsequently, given any subset of examples that wish to be unlearnt, the goal is to learn, without the knowledge of the original training dataset, a good predictor that is identical to the predictor that would have been produced when learning from scratch on the surviving examples.
On User-Level Private Convex Optimization
We introduce a new mechanism for stochastic convex optimization (SCO) with user-level differential privacy guarantees.
Regression with Label Differential Privacy
We study the task of training regression models with the guarantee of label differential privacy (DP).
Private Ad Modeling with DP-SGD
A well-known algorithm in privacy-preserving ML is differentially private stochastic gradient descent (DP-SGD).
Preventing Verbatim Memorization in Language Models Gives a False Sense of Privacy
Studying data memorization in neural language models helps us understand the risks (e. g., to privacy or copyright) associated with models regurgitating training data and aids in the development of countermeasures.
Measuring Forgetting of Memorized Training Examples
In memorization, models overfit specific training examples and become susceptible to privacy attacks.
The Privacy Onion Effect: Memorization is Relative
Machine learning models trained on private datasets have been shown to leak their private data.
Learning to Reason with Neural Networks: Generalization, Unseen Data and Boolean Measures
More generally, the paper considers the learning of logical functions with gradient descent (GD) on neural networks.
Just Fine-tune Twice: Selective Differential Privacy for Large Language Models
Utilizing the fact that sensitive information in language data tends to be sparse, Shi et al. (2021) formalized a DP notion extension called Selective Differential Privacy (SDP) to protect only the sensitive tokens defined by a policy function.
Quantifying Memorization Across Neural Language Models
Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim.
Understanding and Improving Robustness of Vision Transformers through Patch-based Negative Augmentation
We show that patch-based negative augmentation consistently improves robustness of ViTs across a wide set of ImageNet based robustness benchmarks.
Do Vision Transformers See Like Convolutional Neural Networks?
Finally, we study the effect of (pretraining) dataset scale on intermediate features and transfer learning, and conclude with a discussion on connections to new architectures such as the MLP-Mixer.
Pointer Value Retrieval: A new benchmark for understanding the limits of neural network generalization
In PVR, this is done by having one part of the task input act as a pointer, giving instructions on a different input location, which forms the output.
Deduplicating Training Data Makes Language Models Better
As a result, over 1% of the unprompted output of language models trained on these datasets is copied verbatim from the training data.
Understanding Invariance via Feedforward Inversion of Discriminatively Trained Classifiers
We use our reconstruction model as a tool for exploring the nature of representations, including: the influence of model architecture and training objectives (specifically robust losses), the forms of invariance that networks achieve, representational differences between correctly and incorrectly classified images, and the effects of manipulating logits and images.
Deep Learning with Label Differential Privacy
The Randomized Response (RR) algorithm is a classical technique to improve robustness in survey aggregation, and has been widely adopted in applications with differential privacy guarantees.
What is being transferred in transfer learning?
One desired capability for machines is the ability to transfer their knowledge of one domain to another where data is (usually) scarce.
What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation
First, natural image and data distributions are (informally) known to be long-tailed, that is have a significant fraction of rare and atypical examples.
Characterizing Structural Regularities of Labeled Data in Overparameterized Models
We obtain empirical estimates of this score for individual instances in multiple data sets, and we show that the score identifies out-of-distribution and mislabeled examples at one end of the continuum and strongly regular examples at the other end.
Transfusion: Understanding Transfer Learning for Medical Imaging
Investigating the learned representations and features, we find that some of the differences from transfer learning are due to the over-parametrization of standard models rather than sophisticated feature reuse.
Identity Crisis: Memorization and Generalization under Extreme Overparameterization
We study the interplay between memorization and generalization of overparameterized networks in the extreme case of a single training example and an identity-mapping task.
Are All Layers Created Equal?
Morally, layers of large deep neural networks can be categorized as either "robust" or "critical".
Unrestricted Adversarial Examples
We introduce a two-player contest for evaluating the safety and robustness of machine learning systems, with a large prize pool.
A Study on Overfitting in Deep Reinforcement Learning
We conclude with a general discussion on overfitting in RL and a study of the generalization behaviors from the perspective of inductive bias.
Machine Theory of Mind
We design a Theory of Mind neural network -- a ToMnet -- which uses meta-learning to build models of the agents it encounters, from observations of their behaviour alone.
Theory of Deep Learning IIb: Optimization Properties of SGD
In Theory IIb we characterize with a mix of theory and experiments the optimization of deep convolutional networks by Stochastic Gradient Descent.
Understanding deep learning requires rethinking generalization
Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance.
Training Deep Nets with Sublinear Memory Cost
In the extreme case, our analysis also shows that the memory consumption can be reduced to O(log n) with as little as O(n log n) extra cost for forward computation.
MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
This paper describes both the API design and the system implementation of MXNet, and explains how embedding of both symbolic expression and tensor operation is handled in a unified fashion.
Learning with a Wasserstein Loss
In this paper we develop a loss function for multi-label learning, based on the Wasserstein distance.
Learning An Invariant Speech Representation
Recognition of speech, and in particular the ability to generalize and learn from small sets of labelled examples like humans do, depends on an appropriate representation of the acoustic input.
A Deep Representation for Invariance And Music Classification
We present the main theoretical and computational aspects of a framework for unsupervised learning of invariant audio representations, empirically evaluated on music genre classification.
Multi-task Vector Field Learning
MTVFL has the following key properties: (1) the vector fields we learned are close to the gradient fields of the prediction functions; (2) within each task, the vector field is required to be as parallel as possible which is expected to span a low dimensional subspace; (3) the vector fields from all tasks share a low dimensional subspace.
Semi-supervised Regression via Parallel Field Regularization
To achieve this goal, we show that the second order smoothness measures the linearity of the function, and the gradient field of a linear function has to be a parallel vector field.
