Two-layer ReLU networks outperform kernel methods in teacher-student settings.
problem Understanding the excess risk of two-layer ReLU neural networks in teacher-student models.
method Investigated a two-phase training process for a student network, comparing it to kernel methods.
result The student network reaches near-global optimality and outperforms kernel methods in minimax optimal rate.
Study shows how task similarity affects forgetting in teacher-student setup.
problem Catastrophic forgetting in continual learning.
method Extended teacher-student setup to multiple teachers, analyzing similarity between tasks.
result Task similarity, whether at readouts or features, influences forgetting and transfer.
Study analyzes catastrophic forgetting in continual learning using teacher-student networks.
problem Catastrophic forgetting in continuously learning systems.
method Teacher-student learning framework, similarity of input distributions and target functions.
result Network can avoid catastrophic forgetting with small input distribution similarity and large target function similarity.
Paper generalizes teacher-student model for realistic data.
problem Capturing learning curves for realistic datasets.
method Introduces a Gaussian covariate generalization of the teacher-student model.
result Generalized model captures learning curves for various realistic data sets.
Paper analyzes how neural networks learn from a teacher in a specific setting.
problem Understanding how two-layer ReLU neural networks learn from a teacher in a regression model.
method Used gradient descent with specific regularization and over-parameterization, combined with measure representation and sparse estimation.
result Student network can identify teacher network parameters with high probability via gradient descent.
Analyzes Hessian spectrum for neural networks near optimal learning.
problem Understanding learning dynamics near optimal points in neural networks.
method Characterizes Hessian eigenspectrum for teacher-student problems, using analytical and numerical methods.
result The rank of the Hessian matrix determines effective number of parameters for non-linear networks.
Deep convolutional neural networks have been widely used in numerous applications, but their demanding storage and computational resource requirements prevent their applications on mobile devices. Knowledge distillation aims to optimize a portable student network by taking the knowledge from a well-trained heavy teache…
Deep neural networks outperform traditional methods in high-dimensional classification.
problem Understanding the empirical success of deep neural networks in high-dimensional classification.
method Proposed a teacher-student framework with Bayes classifier as ReLU neural networks, derived convergence rates for 0-1 and hinge losses.
result Sharp rate of convergence for classifiers trained using 0-1 or hinge loss, with Od(n−2/3) or Od(n−1) under separable data distribution. High-dimensional data in many machine learning applications leads to computational and analytical complexities. Feature selection provides an effective way for solving these problems by removing irrelevant and redundant features, thus reducing model complexity and improving accuracy and generalization capability of the…
Deep neural networks bring in impressive accuracy in various applications, but the success often relies on the heavy network architecture. Taking well-trained heavy networks as teachers, classical teacher-student learning paradigm aims to learn a student network that is lightweight yet accurate. In this way, a portable…
The paper proposes a Gibbs sampler for neural network posterior sampling.
problem Sampling from the posterior of neural networks.
method Adding noise to activations and using a Gibbs sampler.
result The Gibbs sampler achieves similar performance to MCMC methods on real and synthetic data.
Linear models can grok without understanding, improving generalization.
problem Understanding the phenomenon of grokking in linear models.
method Analytical and numerical derivation of training and generalization dynamics in linear networks.
result Grokking can occur in linear networks without reaching understanding, and its timing depends on various parameters.
Neural networks' performance scales with data size, explained by data manifold dimensionality.
problem Understanding the scaling of neural network performance with the number of parameters.
method Explained by the intrinsic dimension of the data manifold, confirmed through teacher/student framework and various datasets.
result The scaling exponent α is approximately 4 divided by the intrinsic dimension d of the data manifold.
The study analyzes multi-class teacher-student perceptron performance and generalization errors.
problem Analyzing multi-class classification with the teacher-student perceptron.
method Deriving asymptotic expressions for Bayes-optimal and empirical risk minimization (ERM) generalization errors.
result Regularised cross-entropy minimization yields close-to-optimal accuracy for multi-class classification.
Randomly sampled interpolators achieve zero generalization error with enough data.
problem Understanding the high generalization ability of machine learning models.
method Algebraic geometry tools to prove zero generalization error for random interpolators.
result Generalization error of randomly sampled interpolators becomes zero once the number of training samples exceeds a geometric threshold.
Superposition accelerates training to a universal power-law exponent.
problem Training dynamics in neural networks.
method Teacher-student framework and analytic theory.
result Superposition leads to a universal power-law exponent of ~1, independent of data and channel statistics.
Deep neural networks achieve stellar generalisation even when they have enough parameters to easily fit all their training data. We study this phenomenon by analysing the dynamics and the performance of over-parameterised two-layer neural networks in the teacher-student setup, where one network, the student, is trained…
Superpixel-mix enhances reliability in semantic segmentation.
problem Improving reliability in real-world semantic segmentation.
method Superpixel-mix, a new data augmentation method with teacher-student consistency training.
result Superpixel-mix achieves state-of-the-art results in semi-supervised semantic segmentation.
More accurate machine learning models often demand more computation and memory at test time, making them difficult to deploy on CPU- or memory-constrained devices. Teacher-student compression (TSC), also known as distillation, alleviates this burden by training a less expensive student model to mimic the expensive teac…
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.
Self-distillation improves model performance by increasing teacher diversity and smoothing predictions.
problem Improving model generalization and performance through self-distillation.
method Interpreting self-distillation as MAP estimation and proposing instance-specific label smoothing.
result Self-distillation enhances model performance by increasing teacher diversity and smoothing predictions.
A new method for training GNNs without a teacher model.
problem Training over-parameterized GNN models is difficult and inefficient.
method GNN Self-Distillation (GNN-SD) with NDR and ADR.
result Improves GNN performance with less training cost and better generalization.
A new training method improves MLIPs for faster, lighter simulations.
problem High computational and memory costs of complex MLIPs for large-scale MD simulations.
method Teacher-student training framework using latent atomic energy knowledge.
result Lightweight student MLIPs achieve faster MD speeds and comparable accuracy to teachers.
Locality helps in learning from high-dimensional data.
problem Understanding how convolutional neural networks learn from high-dimensional data.
method Teacher-student framework for kernel regression with convolutional kernels.
result Locality is key to determining the learning curve exponent in high-dimensional data.
Deep neural networks undergo hierarchical free-energy landscape transitions with increasing data size.
problem Understanding the design space and dynamics of deep neural networks.
method Statistical mechanical approach based on replica method.
result Hierarchical free-energy landscape transitions with ultrametricity, leading to simpler configurations in deeper layers.
Understanding theoretical properties of deep and locally connected nonlinear network, such as deep convolutional neural network (DCNN), is still a hard problem despite its empirical success. In this paper, we propose a novel theoretical framework for such networks with ReLU nonlinearity. The framework explicitly formul…
Study online learner attacks by manipulating labels, revealing critical thresholds.
problem Adversarial attacks on online learners, especially label manipulation.
method Control-theoretical perspective, theoretical analysis, empirical study.
result Discontinuous transition in learner's accuracy occurs at a critical threshold.
Deep neural networks achieve stellar generalisation on a variety of problems, despite often being large enough to easily fit all their training data. Here we study the generalisation dynamics of two-layer neural networks in a teacher-student setup, where one network, the student, is trained using stochastic gradient de…
Paper presents a deep learning framework for classifying respiratory anomalies and lung diseases from sound recordings.
problem Classifying respiratory anomalies and lung diseases from respiratory sound recordings.
method The framework uses front-end feature extraction to transform sound into spectrograms, and a deep learning network to classify these features.
result The proposed deep learning system outperforms current state-of-the-art methods on the ICBHI benchmark dataset.
A new method identifies a stable subnetwork in overparameterized student models.
problem Identifying a stable subnetwork in overparameterized student models.
method Spectral representation of linear transfer of information, focusing on eigenvalues and eigenvectors.
result A stable student substructure is isolated that mirrors the true complexity of the teacher.
Heuristic tools from statistical physics have been used in the past to locate the phase transitions and compute the optimal learning and generalization errors in the teacher-student scenario in multi-layer neural networks. In this contribution, we provide a rigorous justification of these approaches for a two-layers ne…
For real-world speech recognition applications, noise robustness is still a challenge. In this work, we adopt the teacher-student (T/S) learning technique using a parallel clean and noisy corpus for improving automatic speech recognition (ASR) performance under multimedia noise. On top of that, we apply a logits select…
Vision algorithms capable of interpreting scenes from a real-time video stream are necessary for computer-assisted surgery systems to achieve context-aware behavior. In laparoscopic procedures one particular algorithm needed for such systems is the identification of surgical phases, for which the current state of the a…
Approach for biosensor condition detection using domain adaptation.
problem Limited clinical data for biosensor condition detection.
method Teacher-student adaptation of deep learning models.
result Significantly outperforms other methods on synthetic and real-world datasets.
Finding minima of a real valued non-convex function over a high dimensional space is a major challenge in science. We provide evidence that some such functions that are defined on high dimensional domains have a narrow band of values whose pre-image contains the bulk of its critical points. This is in contrast with the…
This paper explores the use of adversarial examples in training speech recognition systems to increase robustness of deep neural network acoustic models. During training, the fast gradient sign method is used to generate adversarial examples augmenting the original training data. Different from conventional data augmen…
Meta Pseudo Labels boosts image classification accuracy to 90.2%.
problem Improving semi-supervised learning for image classification.
method Adapts a teacher network to generate better pseudo labels through student feedback.
result Achieves a new state-of-the-art top-1 accuracy of 90.2% on ImageNet.
Optimal task order improves continual learning performance.
problem Challenges in neural networks learning multiple tasks in sequence.
method Linear teacher-student model with latent factors, derived analytical expression.
result Two principles for optimal task order: least representative first and dissimilar adjacent tasks.
Recent variants improve knowledge distillation performance.
problem Improving the performance of knowledge distillation.
method Introducing additional components or changing the learning process.
result These variants have shown promising results.
Study shows overparameterization helps shallow neural networks recover signals in high dimensions.
problem Signal recovery in shallow neural networks with overparameterization.
method Gradient flow on population risk, Gaussian distribution assumption, high-dimensional limit analysis.
result Minimal overparameterization is sufficient for strong recovery of signals.
Deep network compression has been achieved notable progress via knowledge distillation, where a teacher-student learning manner is adopted by using predetermined loss. Recently, more focuses have been transferred to employ the adversarial training to minimize the discrepancy between distributions of output from two net…
New insights into neural network forgetting reveal a trade-off between node activation and re-use.
problem Challenges in maintaining performance on old tasks while learning new ones.
method Theoretical analysis of synthetic and real data setups, focusing on node activation vs re-use.
result Worst forgetting occurs in an intermediate similarity regime between learned tasks.
We analyze the dynamics of training deep ReLU networks and their implications on generalization capability. Using a teacher-student setting, we discovered a novel relationship between the gradient received by hidden student nodes and the activations of teacher nodes for deep ReLU networks. With this relationship and th…
A new teacher-class network method compresses DNNs by distributing knowledge to multiple student networks.
problem Overwhelming size of Deep Neural Networks (DNNs).
method Single teacher with multiple student networks, transferring knowledge to each student.
result The combined knowledge of the class of students achieves better performance and reduces parameters.
Study reveals sharp characterisation of local minima in neural network loss landscapes.
problem Characterizing local minima in high-dimensional two-layer ReLU neural networks.
method Exact low-dimensional representation of local minima using summary statistics and link with one-pass SGD dynamics.
result Local minima in overparameterized neural networks form discrete families with varying stability and reachability.
The paper reveals surprising star-shaped connectivity in neural networks.
problem Understanding mode connectivity in neural network landscapes.
method Fine-grained analysis of connectivity in overparameterized and finite minima cases.
result Star-shaped connectivity exists in neural network landscapes, suggesting near convexity.
New statistical mechanics analysis shows edge pruning outperforms node pruning in neural networks.
problem Theoretical understanding of neural network pruning effectiveness is lacking.
method Statistical mechanics analysis of a teacher-student framework.
result DPP node pruning method is superior to other methods, but edge pruning is better overall.
Study on neural network dynamics in high dimensions with quadratic activation.
problem Understanding training dynamics in overparameterized neural networks.
method Derivation of gradient flow equations and analysis under l2-regularization.
result Characterization of estimator performance and spectral properties in the high-dimensional limit.