Student-teacher learning improves generalization with noisy inputs.
problem Transfer knowledge from clean inputs to noisy inputs.
method Analyzes student-teacher learning using deep linear networks and experiments with nonlinear networks.
result Three factors are vital for success: zero training loss, teacher knowledge, and feature decomposition.
Study shows RFRR's effectiveness with nearly orthogonal data in overparameterized settings.
problem Understanding the effectiveness of random feature regression with nearly orthogonal data.
method Investigates RFRR with nearly orthogonal deterministic unit-length input data vectors in the overparameterized regime.
result Shows high-probability non-asymptotic concentration results for RFRR's training, cross-validation, and generalization errors.
Proposes RaT to mitigate bias in student-teacher estimation.
problem Systematic bias in teacher's predictions propagates to student model.
method Uses teacher to estimate residuals in student's predictions.
result RaT method reduces teacher bias effect and achieves optimal rate.
Self-training with noisy student-teacher boosts keyword spotting accuracy.
problem Robust keyword spotting in challenging conditions.
method Aggressive data augmentation and self-training with noisy student-teacher approach.
result Significant accuracy improvement in difficult conditions, up to 60%.
A new method detects concept drift without true labels.
problem Detecting concept drift in unsupervised settings.
method Student-teacher learning paradigm for drift detection.
result The method outperforms state-of-the-art approaches in experiments.
LoRA fine-tuning explained with gradient dynamics for low-rank perturbations.
problem Understanding why gradient descent converges to useful low-rank perturbations in LoRA fine-tuning.
method Generalized student-teacher setting with i.i.d. samples and online gradient descent.
result Gradient descent converges to the teacher model in dkO(1) iterations under certain conditions. KD can lead to student-teacher deviations that improve performance.
problem KD can lead to student-teacher deviations that may outperform the teacher.
method Characterized and explained the nature of student-teacher deviations through experiments and theory.
result KD can lead to improved generalization by exaggerating the implicit bias of gradient descent.
This work proposes a student-teacher network for predicting hospital admission locations.
problem Accurate prediction of hospital admission locations to optimize resource allocation.
method Reinforcement learning approach where a teacher network selects data batches for a student network.
result The approach outperforms state-of-the-art methods on tabular data and image recognition.
Study analyzes learning dynamics in nonlinear perceptrons using stochastic-process approach.
problem Understanding the roles of nonlinearity and input-data distribution in neural network learning.
method Stochastic-process approach to derive flow equations for learning dynamics in nonlinear perceptrons.
result Input-data noise affects learning speed differently under supervised and reinforcement learning.
Framework models supervised learning in non-stationary data.
problem Non-stationary data in supervised learning.
method Statistical physics methods applied to LVQ and neural networks.
result LVQ and ReLU have different sensitivity to concept drift.
Study analyzes Bayesian inference algorithms using dynamical functional approach.
problem Analysis of approximate inference algorithms for large Gaussian latent variable models.
method Dynamical functional approach to model nontrivial dependencies and obtain exact effective stochastic process.
result Closed-form expressions for the rate of convergence are derived and validated.
This is a report of our lessons learned building acoustic models from 1 Million hours of unlabeled speech, while labeled speech is restricted to 7,000 hours. We employ student/teacher training on unlabeled data, helping scale out target generation in comparison to confidence model based methods, which require a decoder…
Large scale machine learning (ML) systems such as the Alexa automatic speech recognition (ASR) system continue to improve with increasing amounts of manually transcribed training data. Instead of scaling manual transcription to impractical levels, we utilize semi-supervised learning (SSL) to learn acoustic models (AM) …
Lifelong learning is the problem of learning multiple consecutive tasks in a sequential manner, where knowledge gained from previous tasks is retained and used to aid future learning over the lifetime of the learner. It is essential towards the development of intelligent machines that can adapt to their surroundings. I…
We study layered neural networks of rectified linear units (ReLU) in a modelling framework for stochastic training processes. The comparison with sigmoidal activation functions is in the center of interest. We compute typical learning curves for shallow networks with K hidden units in matching student teacher scenarios…
BootsTAP uses real-world data to improve TAP tracking performance.
problem Lack of large-scale groundtruth training data for TAP.
method Bootstrapped training with a self-supervised student-teacher setup.
result State-of-the-art performance on TAP-Vid benchmarks.
Although reinforcement learning (RL) can provide reliable solutions in many settings, practitioners are often wary of the discrepancies between the RL solution and their status quo procedures. Therefore, they may be reluctant to adapt to the novel way of executing tasks proposed by RL. On the other hand, many real-worl…
Overcoming the visual barrier and developing "see-through vision" has been one of mankind's long-standing dreams. Unlike visible light, Radio Frequency (RF) signals penetrate opaque obstructions and reflect highly off humans. This paper establishes a deep-learning model that can be trained to reconstruct continuous vid…
A new gradient flow for MMD with closed-form implementation.
problem Existing gradient flows either lack tractable numerical implementation or require strong assumptions.
method Introduces a (de)-regularized Maximum Mean Discrepancy (DrMMD) and its gradient flow.
result Guarantees near-global convergence for a broad class of targets in both continuous and discrete time.
User-generated reviews can be decomposed into fine-grained segments (e.g., sentences, clauses), each evaluating a different aspect of the principal entity (e.g., price, quality, appearance). Automatically detecting these aspects can be useful for both users and downstream opinion mining applications. Current supervised…
Knowledge distillation (KD) is a popular method for reducing the computational overhead of deep network inference, in which the output of a teacher model is used to train a smaller, faster student model. Hint training (i.e., FitNets) extends KD by regressing a student model's intermediate representation to a teacher mo…
We find a convex model for traditional nonlinear regression under L2 loss.
problem Nonlinear regression under L2 loss with non-convex optimization.
method Showed a convex nonlinear regression model for least squares problem.
result Existence of a convex model simplifies training complex systems.
Extends importance sampling to nonlinear models using adjoint operators.
problem Lack of tools for identifying important data points in nonlinear models.
method Introduces adjoint operator for nonlinear maps, generalizes norm and leverage scores.
result Generalized scores provide approximation guarantees for nonlinear mappings.
Machine learning maps knots to embeddings, revealing topological invariants.
problem Learning topological invariance in knot theory.
method Contrastive and generative machine learning techniques, auto-regressive decoder Transformer network.
result Neural networks can map different knots to the same point in an embedding vector space.
Unified analysis for nonlinear parametric models in Bayesian optimization.
problem Limited theoretical guarantees for nonlinear parametric models in Bayesian optimization.
method Kernel-based framework for analyzing regularized nonlinear parametric models trained on adaptively collected data.
result Unified convergence guarantees for nonlinear acquisition and surrogate models.
The paper introduces a fast algorithm for learning and forecasting nonlinear dynamics from noisy time series data.
problem Challenges in capturing nonlinear dynamics from noisy time series data.
method A projected nonlinear state-space model with kernel functions applied to projected lines.
result The model effectively learns and forecasts complex nonlinear dynamics with computational efficiency.
AdaKoop efficiently models nonlinear dynamics from nonstationary data streams.
problem Capturing nonlinear dynamics in nonstationary data streams with computational efficiency.
method Koopman operator theory and probabilistic framework for streaming data.
result AdaKoop outperforms state-of-the-art methods in real-time forecasting accuracy and efficiency.
Optimistic estimate predicts best fitting performance of nonlinear models.
problem Evaluating the potential of nonlinear models in fitting.
method Proposes an optimistic estimate to quantify the smallest sample size for fitting nonlinear models.
result Predicts specific subsets of targets that can be fitted at overparameterization.
PETAL adapts models to changing target domains over time.
problem Lifelong test-time adaptation in changing target domains.
method Probabilistic framework with student-teacher model and data-driven parameter restoration.
result PETAL achieves better results than state-of-the-art for online lifelong test-time adaptation.
Causal Mosaic distinguishes cause from effect using nonlinear ICA and ensemble methods.
problem Distinguishing cause from effect in bivariate settings.
method Nonlinear ICA and ensemble framework (Causal Mosaic).
result Causal Mosaic shows state-of-the-art performance on artificial and real-world datasets.
The paper develops adaptive deep learning methods for nonlinear time series models.
problem Estimating mean functions of non-stationary and nonlinear time series models.
method Develops non-penalized and sparse-penalized DNN estimators for general non-stationary time series, derives minimax lower bounds, and shows the sparse-penalized DNN estimator is adaptive and optimal.
result Sparse-penalized DNN estimator achieves minimax optimal rates for many nonlinear AR models.
Study on signal-plus-noise decomposition in nonlinear spiked random matrices.
problem Nonlinear spiked random matrix models with rank-one signal and noise.
method Signal-plus-noise decomposition and phase transition analysis.
result Identified precise phase transitions in signal components at critical thresholds.
Bayesian filtering approach identifies nonlinear restoring forces in dynamic systems.
problem Identification of nonlinear dynamic systems in engineering.
method Modeling the nonlinear restoring force as a Gaussian process, converting it to a state-space model, and inferring internal states and the nonlinear restoring force through filtering and smoothing.
result The approach effectively identifies nonlinear restoring forces in both simulated and experimental datasets.
Characterizes Hessian eigenspectra for realistic nonlinear models.
problem Understanding Hessian eigenspectra in realistic nonlinear models.
method Deterministic equivalent techniques from random matrix theory.
result Hessian can have qualitatively different spectral behaviors.
Optimized nonlinearities enhance generalization in random feature models.
problem Improving generalization performance in random feature models.
method Analyzed and defined optimal nonlinearities from Gaussian model parameters.
result Optimized nonlinearities achieve better generalization performance than ReLU.
Stochastic VB improves nonlinear model inference speed and accuracy.
problem Bayesian inference of nonlinear models from noisy data.
method Stochastic Variational Bayesian (VB) inference for nonlinear models.
result Stochastic VB achieves comparable parameter recovery to analytical solution but is faster.
The paper introduces reservoir computing models for complex systems.
problem Modeling complex engineering systems using nonlinear autoregression.
method Introduces reservoir computing with output feedback as stationary and ergodic infinite-order nonlinear autoregressive models.
result Demonstrates versatility of classical and quantum reservoir computers in modeling synthetic and real data.
Two new Koopman models improve nonlinear system prediction.
problem Predicting nonlinear, nonconvex dynamic systems.
method Convex and Extended Koopman Models using deep learning.
result Significantly improved predictive performance.
Develops a new method for nonlinear dimension reduction using random features.
problem Statistical challenges in generalizing Gaussian process-based latent variable models to non-Gaussian data.
method Random feature latent variable models (RFLVMs) that approximate nonlinear relationships with linear functions of random features.
result RFLVMs produce comparable results to state-of-the-art methods on various data types.
JULIA combines multi-linear and nonlinear models for tensor completion.
problem Complex patterns in real-world tensors require a unified model.
method JULIA unifies multi-linear and nonlinear models with flexible component assignment and efficient alternating optimization.
result JULIA outperforms existing methods in large-scale tensor completion.
Bayesian method improves predictions in overparameterized nonlinear regression.
problem Understanding overparameterization in nonlinear regression models.
method Bayesian framework with adaptive prior considering data spectral structure.
result Posterior contraction established for generalized linear and single-neuron models, demonstrating prediction consistency.
New metric improves latent dynamics inference from neural data.
problem Limitations of co-smoothing in predicting latent dynamics.
method Few-shot co-smoothing to assess latent dynamics.
result High co-smoothing models often have extraneous dynamics, which few-shot co-smoothing detects.
This research highlights the secrecy potential of nonlinear generative models and their all-or-nothing phase transition.
problem Secrecy potential of nonlinear generative models in statistical learning.
method Replica method to derive asymptotic normalized cross entropy and statistical decoupling of Bayesian estimator.
result Strictly nonlinear models exhibit an all-or-nothing phase transition, leading to perfect secrecy.
Linear principal component analysis (PCA) can be extended to a nonlinear PCA by using artificial neural networks. But the benefit of curved components requires a careful control of the model complexity. Moreover, standard techniques for model selection, including cross-validation and more generally the use of an indepe…
WICA improves ICA results with a new method.
problem Finding independent components in nonlinear data.
method New nonlinear ICA model (WICA) with efficient correlation coefficient verification.
result WICA yields better and more stable results than other algorithms.
We introduce the truncated Gaussian graphical model (TGGM) as a novel framework for designing statistical models for nonlinear learning. A TGGM is a Gaussian graphical model (GGM) with a subset of variables truncated to be nonnegative. The truncated variables are assumed latent and integrated out to induce a marginal m…
New framework IIA identifies innovations in general nonlinear vector autoregressive processes.
problem Limited generality of NVAR models due to additive innovation assumption.
method Independent Innovation Analysis (IIA) framework, assuming mutual independence and modulation by an auxiliary variable.
result Guarantees identifiability of innovations with arbitrary nonlinearities, up to permutation and component-wise invertible nonlinearities.
We develop a new DTSM with nonlinearities using Gaussian Processes for better interest rate forecasting.
problem Linear DTSMs fail to capture nonlinear relationships between macroeconomic variables and interest rates.
method We propose a Gaussian Process-based sequential Monte Carlo estimation and forecasting scheme.
result Nonlinear models outperform linear ones in forecasting core inflation, leading to significant economic value gains.