Near-optimal rates for multi-task learning with shared representations.
problem Approximation and statistical complexity of learning multiple operators.
method Multiple Neural Operators (MNO) architecture and comparison with DeepONet.
result Near-optimal upper and lower bounds for approximation and generalization.
Bayesian framework for sequential learning tasks with low-rank approximations.
problem Balancing knowledge retention and adaptability in sequential neural networks.
method Bayesian framework with diagonal plus low-rank approximations of the precision matrix.
result Unlocking capabilities to encode task relationships and incorporate prior knowledge from later tasks.
This work improves multi-task regression performance using approximations of full-conformal prediction.
problem Jointly solve multiple regression tasks with better performance than individual tasks.
method Design an approximating prediction region in RKHS for full-conformal multi-task regression.
result The approximating prediction region improves upon split-conformal prediction on synthetic data.
Transformers enable in-context learning with guarantees for a wide range of tasks.
problem How to enable in-context learning with transformers for various tasks.
method Developed a universal approximation theory integrating Barron's function approximation with transformer capabilities.
result Transformers can approximate any target function with vanishingly small risk using a few in-context examples.
Extends neural network approximations to guarantee continuity of real-world learning tasks.
problem Guaranteeing continuity of real-world learning tasks given by conditional expectations.
method Establishing conditions on learning tasks that guarantee their continuity under a factorization of the data-generating process.
result Conditions guaranteeing the continuity of practically any derived learning task.
Meta-learning variance reduced via Laplace approximation for regression tasks.
problem High variance in meta-learning due to limited support data for each task.
method Laplace approximation to estimate posterior variance and reduce gradient estimate variance.
result Effective variance reduction in meta-learning, improving generalization performance.
The ability of a reinforcement learning (RL) agent to learn about many reward functions at the same time has many potential benefits, such as the decomposition of complex tasks into simpler ones, the exchange of information between tasks, and the reuse of skills. We focus on one aspect in particular, namely the ability…
Improved continual learning for neural networks with BN layers using K-FAC extension.
problem Continual learning challenges in neural networks with BN layers.
method Extended K-FAC method to account for inter-example relations, weight merging, and reparameterization for BN layers; proposed weight merging and reparameterization for BN layers; proposed method to select hyperparameters without source task data.
result Better performance in continual learning tasks with BN layers compared to baselines.
Paper extends multi-task Gaussian Cox processes for heterogeneous tasks.
problem Modeling multiple heterogeneous correlated tasks jointly.
method Data augmentation and mean-field approximation for non-conjugate Bayesian inference.
result Demonstrates improved performance and inference on synthetic and real data.
Study compares adaptive vs fixed query learning methods.
problem Comparing adaptive and fixed query learning methods for task approximation.
method Examined in-context and agentic learning in two settings: unrestricted and realizable.
result Adaptivity does not hinder performance in unrestricted setting but can in realizable setting.
Bayesian unlearning uses Bayes' rule to remove data from a model, but faces challenges in obtaining the exact posterior.
problem Removing data from a trained model while maintaining model accuracy.
method Uses Laplace approximation and Variational Inference to approximate the updated posterior.
result Insights on the applicability of Bayesian unlearning in practical scenarios for neural networks.
A major challenge in reinforcement learning (RL) is the design of agents that are able to generalize across tasks that share common dynamics. A viable solution is meta-reinforcement learning, which identifies common structures among past tasks to be then generalized to new tasks (meta-test). In meta-training, the RL ag…
Convolutional networks outperform shallow classifiers on certain tasks due to hierarchical structure.
problem Understanding why convolutional networks outperform shallow classifiers on specific tasks.
method Approximation theory, visual tasks with deterministic scrambling, and network performance evaluation.
result Hierarchical structure is crucial for convolutional networks' performance on certain tasks, but not all.
A Boolean algebra formalizes task composition for reinforcement learning.
problem Formalizing task composition for efficient learning and problem-solving.
method Formalized tasks as a Boolean algebra, learning goal-oriented value functions, and composing them to solve new tasks.
result Agents can solve new tasks without additional learning by composing value functions in specific ways.
This paper tackles catastrophic forgetting in neural networks by providing a unified framework for regularization-based continual learning.
problem Catastrophic forgetting in neural networks trained sequentially on multiple tasks.
method Formulates regularization-based continual learning as a second-order Taylor approximation of the loss function, leading to a unified framework.
result Theoretical results indicate the importance of accurate approximation of the Hessian matrix for optimization and generalization.
Approximate dynamic programming algorithms, such as approximate value iteration, have been successfully applied to many complex reinforcement learning tasks, and a better approximate dynamic programming algorithm is expected to further extend the applicability of reinforcement learning to various tasks. In this paper w…
A framework isolates and learns approximately shared features for better domain adaptation.
problem Reducing performance degradation in unseen domains using machine learning models.
method Statistical framework distinguishing feature utilities based on correlation variance across domains. Learning approximately shared features from source tasks and fine-tuning on target tasks.
result Improved population risk compared to previous results on both source and target tasks, resolving the paradox of feature selection.
In many areas of machine learning, it becomes necessary to find the eigenvector decompositions of large matrices. We discuss two methods for reducing the computational burden of spectral decompositions: the more venerable Nystom extension and a newly introduced algorithm based on random projections. Previous work has c…
It is widely accepted that optimization of medical imaging system performance should be guided by task-based measures of image quality (IQ). Task-based measures of IQ quantify the ability of an observer to perform a specific task such as detection or estimation of a signal (e.g., a tumor). For binary signal detection t…
Learning new tasks with few samples using related task evaluations.
problem Learning a new task with limited data and related task evaluations.
method Modeling task relatedness through weak monotonicity and leveraging it in transfer learning and model selection aggregation.
result Pruning the model class based on monotonicity and hedging on the task frontier.
Bayesian meta-reinforcement learning improves over point estimates with Laplace approximation.
problem Improving meta-reinforcement learning by providing full posterior distributions.
method Augmenting point estimates with Laplace approximation for full posterior distributions.
result Our method performs similarly to variational baselines with fewer parameters.
Meta-learn Bayesian inference for task-specific BNNs using amortised inference.
problem Efficiently learning Bayesian inference for small-scale probabilistic meta-learning.
method Replace global inducing points with actual data to create a set of approximate likelihoods, train a meta-model to learn these parameters across related datasets.
result Meta-learned inference can be applied to task-specific BNNs, improving efficiency and scalability.
New algorithm estimates task affinities without repeated training, improving model performance and efficiency.
problem Efficiently estimating task affinities among multiple tasks for model training.
method Grad-TAG algorithm: trains a base model for all tasks and uses gradient-based linearization to estimate task affinities.
result Estimates task affinities with high accuracy and low computational cost.
Random features approach has been widely used for kernel approximation in large-scale machine learning. A number of recent studies have explored data-dependent sampling of features, modifying the stochastic oracle from which random features are sampled. While proposed techniques in this realm improve the approximation,…
Pseudo-Likelihood Inference improves ABC for high-dimensional Bayesian inference.
problem Intractable likelihood in Bayesian system identification.
method PLI combines neural approximation with integral probability metrics and adaptive bandwidth.
result PLI outperforms SNPE on challenging tasks, especially with more data.
New method accelerates CNNs for mobile devices by approximating tensors and quantizing weights.
problem Efficiently compress and accelerate CNNs for mobile devices.
method Low-rank tensor approximation in Tucker format combined with quantization of weights and activations.
result Our method significantly improves CNN performance on various classification tasks.
Paper develops a theory explaining contrastive pre-training for multimodal AI.
problem Limited theoretical understanding of contrastive pre-training for multi-modal AI.
method Introduces approximate sufficient statistics and Joint Generative Hierarchical Model.
result Near-minimizers of contrastive loss are approximately sufficient, enabling diverse downstream tasks.
Collaborative filtering (CF) is a popular technique in today's recommender systems, and matrix approximation-based CF methods have achieved great success in both rating prediction and top-N recommendation tasks. However, real-world user-item rating matrices are typically sparse, incomplete and noisy, which introduce ch…
We consider evidence integration from potentially dependent observation processes under varying spatio-temporal sampling resolutions and noise levels. We develop a multi-resolution multi-task (MRGP) framework while allowing for both inter-task and intra-task multi-resolution and multi-fidelity. We develop shallow Gauss…
One of the central goals of Recurrent Neural Networks (RNNs) is to learn long-term dependencies in sequential data. Nevertheless, the most popular training method, Truncated Backpropagation through Time (TBPTT), categorically forbids learning dependencies beyond the truncation horizon. In contrast, the online training …
New algorithm improves knowledge transfer in dynamic decision-making.
problem Utilizing data from existing ventures to improve decision-making in new ventures.
method Proposes Transferred Fitted Q-Iteration algorithm for estimating optimal action-state function Q∗. result Significantly improved final learning error of Q∗ function. In this thesis I explore challenging discrete energy minimization problems that arise mainly in the context of computer vision tasks. This work motivates the use of such "hard-to-optimize" non-submodular functionals, and proposes methods and algorithms to cope with the NP-hardness of their optimization. Consequently, t…
Unified view of federated learning and distributed RL using local stochastic approximation.
problem Finding the root of an operator composed of local operators in a network of agents with dependent data.
method Local stochastic approximation over a network of agents with Markov process-dependent data.
result Convergence rates of local stochastic approximation for both constant and time-varying step sizes, within a logarithmic factor of independent data.
High-dimensional observations and complex real-world dynamics present major challenges in reinforcement learning for both function approximation and exploration. We address both of these challenges with two complementary techniques: First, we develop a gradient-boosting style, non-parametric function approximator for l…
Proposes a Quasi-Newton trust region method for policy optimization in reinforcement learning.
problem Lack of stepsize selection criterion and slow convergence in gradient descent for policy optimization.
method Uses a trust region method with Quasi-Newton approximation for the Hessian.
result Demonstrates improved performance and efficiency in continuous control tasks.
One of the most challenging problems in kernel online learning is to bound the model size and to promote the model sparsity. Sparse models not only improve computation and memory usage, but also enhance the generalization capacity, a principle that concurs with the law of parsimony. However, inappropriate sparsity mode…
Neural networks help create summary statistics for complex models.
problem Creating summary statistics for models with intractable likelihood functions.
method Infomax learning with neural networks to maximize mutual information.
result Improves performance of approximate Bayesian computation and neural likelihood methods.
Extreme learning machine (ELM) is a new single hidden layer feedback neural network. The weights of the input layer and the biases of neurons in hidden layer are randomly generated, the weights of the output layer can be analytically determined. ELM has been achieved good results for a large number of classification ta…
Paper presents a deterministic method for diverse subset selection.
problem Diverse subset selection problems in recommendation, summarization, and search.
method Greedy deterministic adaptation of k-DPP for low-rank approximations and image search.
result The method yields low-rank approximations of kernel matrices and demonstrates effectiveness in image search.
Paper explores supervised learning methods to approximate ideal observer for joint signal detection and localization.
problem Optimizing medical imaging systems by assessing their performance using the Ideal Observer model.
method Uses supervised learning methods, specifically convolutional neural networks, to approximate the Ideal Observer for joint signal detection and localization tasks.
result Supervised learning-based methods can approximate the Ideal Observer for joint signal detection and localization tasks, as shown by comparisons to MCMC and analytical methods.
TDS provides exact samples for conditional distributions in diffusion models.
problem Lack of exact sampling methods for diffusion models.
method Sequential Monte Carlo (SMC) algorithm with twisting technique.
result TDS offers more accurate approximations with fewer particles compared to heuristics.
FastDTW is often slower than exact DTW in practical applications.
problem The performance gap between FastDTW and exact DTW in real-world data mining tasks.
method Comparison of FastDTW and exact DTW performance in various data mining applications.
result FastDTW is generally slower than exact DTW in practical applications.
Current approaches in approximate inference for Bayesian neural networks minimise the Kullback-Leibler divergence to approximate the true posterior over the weights. However, this approximation is without knowledge of the final application, and therefore cannot guarantee optimal predictions for a given task. To make mo…
Proposes a new method to approximate Gaussian inference in classification tasks.
problem Uncertainty quantification in classification tasks using softmax functions.
method Develops a new formalism to approximate Gaussian distributions over logit space and proposes using element-wise normCDF or sigmoid instead of softmax.
result Improves uncertainty quantification compared to softmax Monte Carlo sampling.
Develops a scalable multi-task Gaussian process with neural embedding for improved performance.
problem High model complexity and limited model capability in multi-task Gaussian processes.
method Neural embedding of coregionalization, advanced variational inference, sparse approximation.
result Higher prediction quality and better generalization of the NSVLMC model.
MoEs can efficiently model complex tasks with low-dimensionality and sparsity.
problem Understanding the theoretical foundations of MoEs for complex tasks.
method Systematic study of MoEs with two structural priors: low-dimensionality and sparsity.
result MoEs can approximate functions on low-dimensional manifolds and exhibit exponential structured tasks.
Meta-learning reformulated as Bayesian risk minimization.
problem Learning models to quickly adapt to new tasks from small datasets.
method Formalized meta-learning as Bayesian risk minimization, using a probabilistic framework to compute predictive distributions from posterior distributions of latent variables conditioned on contextual datasets.
result A novel Gaussian approximation for the posterior distribution that converges to maximum likelihood estimates and outperforms Neural Process on benchmark datasets.
HyperX uses reward bonuses to enable efficient exploration in meta-learning.
problem Catastrophic failure of meta-learning with sparse rewards.
method HyperX uses novel reward bonuses to explore in approximate hyper-state space.
result HyperX meta-learns better task-exploration and adapts more successfully to new tasks.