Paper defines a new distance metric for comparing learning tasks.
problem Comparing difficulty of learning tasks between source and target.
method Information geometry, optimal transport, coupled transfer distance.
result Coupled transfer distance correlates with fine-tuning difficulty.
Method uses random forest with distance covariance for transfer learning in healthcare.
problem Transfer learning in random forests with sparse differences between source and target.
method Distance covariance-based feature weights in residual random forest.
result Upper bound on mean square error rate for transfer learning in RF.
WAPPO optimizes feature distributions for better visual transfer in RL.
problem Improving visual transfer in reinforcement learning.
method WAPPO uses Wasserstein Confusion to minimize feature distribution distance.
result WAPPO outperforms previous methods in visual transfer across different environments.
New study reveals surprising adaptive rates in model selection for transfer learning.
problem Model selection in transfer learning with varying transfer distances.
method Theoretical analysis focusing on classification.
result Adaptive rates can be arbitrarily slower than oracle rates.
The goal of transfer learning is to improve the performance of target learning task by leveraging information (or transferring knowledge) from other related tasks. In this paper, we examine the problem of transfer distance metric learning (DML), which usually aims to mitigate the label information deficiency issue in t…
Paper proposes a new Wasserstein distance for mixtures of radially contoured distributions.
problem Generalization of Wasserstein distance to non-elliptically contoured distributions.
method Relaxed formulation for mixtures of radially contoured distributions without marginal consistency.
result The new distance yields more stable error and better color distribution in image transfer tasks.
Optimal transport metric transfers deep network representations efficiently.
problem Efficiently transfer deep network representations for new tasks.
method Use optimal transport to quantify representation similarity and regularize student network.
result Optimal transport distance promotes similarity between teacher and student representations.
The demand of artificial intelligent adoption for condition-based maintenance strategy is astonishingly increased over the past few years. Intelligent fault diagnosis is one critical topic of maintenance solution for mechanical systems. Deep learning models, such as convolutional neural networks (CNNs), have been succe…
A new framework for robust transfer learning that avoids negative transfer in domains with unequal information.
problem Negative transfer in unsupervised domain adaptation, especially when source and target domains have different levels of informativeness.
method Decision-theoretic framework based on Le Cam's theory of statistical experiments, using constructive approximations to replace strict invariance with directional simulability.
result Le Cam Distortion achieves near-perfect frequency estimation and zero source utility loss in various domains, demonstrating superior performance compared to traditional methods.
A new dataset distance using optimal transport, agnostic of model and label sets.
problem Quantifying task similarity across datasets without model dependence.
method Optimal transport for model-agnostic dataset comparison.
result The new distance correlates with transfer learning difficulty across various datasets.
Paper proposes dp-VAE for preserving spatial context in gene expression data.
problem Inaccessibility of spatial context in single-cell gene expression data.
method Generic representation learning and transfer learning framework with a distance-preserving regularizer.
result dp-VAE effectively reconstructs and imputes spatial context from gene expression data.
Paper introduces new Gromov-type distances for comparing Gaussian mixture models.
problem Comparing distributions across different metric spaces using Gromov-Wasserstein distances.
method Incorporates invariance properties into MW2, introducing MGW2 and EW2.
result MGW2 and EW2 are efficient for estimating distances between GMMs in practical applications.
We propose a novel end-to-end non-minimax algorithm for training optimal transport mappings for the quadratic cost (Wasserstein-2 distance). The algorithm uses input convex neural networks and a cycle-consistency regularization to approximate Wasserstein-2 distance. In contrast to popular entropic and quadratic regular…
For a pinched Hadamard manifold X and a discrete group of isometries Γ of X, the critical exponent δΓ is the exponential growth rate of the orbit of a point in X under the action of Γ. We show that the critical exponent for any family N of normal subgroups of Γ0 has the same coarse behaviour…
Generative model downgrades coarse satellite images to fine resolution.
problem Reconstructing fine resolution satellite images from coarse scale inputs.
method Combines U-Net transfer encoder with diffusion-based generative model.
result Excellent performance (R2 = 0.65 to 0.94) across seasonal regional splits.
Distance metric learning (DML) aims to find an appropriate way to reveal the underlying data relationship. It is critical in many machine learning, pattern recognition and data mining algorithms, and usually require large amount of label information (such as class labels or pair/triplet constraints) to achieve satisfac…
Proposes an energy-based sliced Wasserstein distance for improved probability measure comparison.
problem Inefficiencies and limitations in existing sliced Wasserstein distance approaches.
method Introduces an energy-based slicing distribution for better performance and stability.
result Demonstrates superior performance of the EBSW distance in various applications.
Study geometric structures in transfer learning to avoid negative transfer.
problem Understanding information-theoretic limits of transfer learning without exploiting domain geometry.
method Integrates geometric structure into linear regression models, using Gram matrices of source and target domains.
result Proposes an interpolation estimator that matches minimax lower bound and outperforms existing methods.
Two Anosov metrics with same boundary distance are isometric.
problem Boundary rigidity for surfaces of Anosov type.
method Transfer principle linking marked length spectrum rigidity to marked boundary distance rigidity.
result Two metrics of Anosov type with the same marked boundary distance are isometric.
The ability of a human being to extrapolate previously gained knowledge to other domains inspired a new family of methods in machine learning called transfer learning. Transfer learning is often based on the assumption that objects in both target and source domains share some common feature and/or data space. In this p…
s-OTDD compares datasets efficiently without training, robust to class variations.
problem Efficiently compare datasets without training or class variations.
method Moment Transform Projection (MTP) and sliced optimal transport.
result s-OTDD correlates with optimal transport and transfer learning performance.
Paper explores properties of slice-matching operators for measure transfer.
problem Efficiently transferring measures in high dimensions.
method Examines an associated slice-matching operator with source, target measures and slicing directions.
result Establishes invariance, equivariance, Lipschitz continuity, and error bounds.
Study improves hypothesis transfer learning for functional linear models.
problem Incompatible TL techniques for high-dimensional FLR methods due to infinite-dimensional nature of functional data.
method Proposes two algorithms for hypothesis transfer learning in RKHS framework, leveraging RKHS distance and aggregation techniques.
result Establishes asymptotic lower bounds and matching upper bounds for the proposed algorithms, demonstrating their effectiveness.
Grogan et al [11,12] have recently proposed a solution to colour transfer by minimising the Euclidean distance L2 between two probability density functions capturing the colour distributions of two images (palette and target). It was shown to be very competitive to alternative solutions based on Optimal Transport for c…
This paper defines and quantifies transferability in domain generalization.
problem Understanding and quantifying transferability between domains.
method Formal definition and estimation of transferability, upper bound for target error.
result Many algorithms do not learn transferable features, proposing a new algorithm.
Paper proves convergence of measure transfer schemes using slicing and matching.
problem Iterative schemes for measure transfer and approximation problems.
method Slicing-and-matching procedure, stochastic gradient descent on Wasserstein space.
result Almost sure convergence proof for stochastic slicing-and-matching schemes.
Improves robust transfer learning with side information.
problem Addressing environmental shift in MDPs with side information.
method Estimate-centered uncertainty sets with side information integration.
result Improved robust policy with reduced sub-optimality gap.
Landmark-based node embeddings approximate shortest path distances in random graphs.
problem Capturing global graph distances in node representations.
method Landmark-based node embeddings using shortest path distances from a subset of reference nodes (landmarks).
result Random graphs require lower dimensions in landmark-based embeddings compared to worst-case graphs.
ATD measures language distance using neural models, recovering linguistic groupings.
problem Lack of a unified quantitative measure for cross-linguistic distance.
method Pretrained multilingual language models, attention mechanisms, optimal transport.
result ATD quantifies representational distance between languages, recovering linguistic groupings.
Trans-GLMC tackles source heterogeneity in transfer learning for structured clusters.
problem Source heterogeneity makes it hard to use multiple related auxiliary sources effectively.
method Trans-GLMC constructs clusters of sources, then combines global fusion, within-cluster refinement, and target debiasing.
result Improves facility-specific prediction and identifies interpretable communities of hospitals with mutual transferability.
Introduces MSW distances to improve SW metrics.
problem Redundant projections in SW distance.
method Imposes Markov structure on projecting directions.
result MSW distances improve SW metrics.
A powerful network teaches a weak one, improving its performance.
problem Improving the performance of a weak neural network using a more powerful one.
method During training, a weak network learns features from a strong network to minimize feature distance.
result A weak neural network can increase its performance by learning from a more powerful network.
This paper analyzes minibatch optimal transport distances and their applications.
problem Optimal transport distances are complex and impractical for large datasets.
method Extended analysis of minibatch optimal transport distances, focusing on various kernels and debiased functions.
result Minibatch optimal transport distances are unbiased estimators and have statistical and optimisation properties.
Background: Pharmacokinetic evaluation is one of the key processes in drug discovery and development. However, current absorption, distribution, metabolism, excretion prediction models still have limited accuracy. Aim: This study aims to construct an integrated transfer learning and multitask learning approach for deve…
Transfer learning has achieved promising results by leveraging knowledge from the source domain to annotate the target domain which has few or none labels. Existing methods often seek to minimize the distribution divergence between domains, such as the marginal distribution, the conditional distribution or both. Howeve…
New Gromov-Wasserstein metric controls rigidity and incorporates prior knowledge.
problem Inflexible Gromov-Wasserstein distance and lack of feature alignment.
method Augmented Gromov-Wasserstein distance with feature alignments and prior knowledge.
result Improved performance in single-cell multi-omic alignment and transfer learning.
We compute the transition probability between two learning tasks, and show that it decomposes into two factors. The first depends on the geometry of the loss landscape of a model trained on each task, independent of any particular model used. This is related to an information theoretic distance function, but is insuffi…
We explain how the Transference Principles from Diophantine approximation can be interpreted in terms of geometry of the locally symmetric spaces Tn=SO(n)\SL(n,R)/SL(n,Z) with n>1, and how, via this dictionary, they become transparent geometric remarks and can be easily proved. Indeed, a finite family …
Optimal transport distances are powerful tools to compare probability distributions and have found many applications in machine learning. Yet their algorithmic complexity prevents their direct use on large scale datasets. To overcome this challenge, practitioners compute these distances on minibatches {\em i.e.} they a…
New distance metric for neural architecture search reduces search space complexity.
problem Reducing the complexity of neural architecture search.
method Fisher task distance for measuring task similarity and online neural architecture search.
result Reduced search space complexity for task-specific architectures.
A new method constrains deep networks during fine-tuning to improve generalization.
problem Improving generalization of fine-tuned deep networks.
method A neural network generalisation bound based on distance from initial weights constrains the hypothesis class to a small sphere.
result Empirical evaluation shows superior generalization performance compared to existing methods.
This work investigates how neural collapse improves transfer learning for large-scale models.
problem Improving transfer learning for large-scale models with limited labeled data.
method Investigates neural collapse and develops a fine-tuning method using skip-connections.
result Feature collapse on downstream data correlates with higher transfer accuracy.
Word embeddings, i.e., low-dimensional vector representations such as GloVe and SGNS, encode word "meaning" in the sense that distances between words' vectors correspond to their semantic proximity. This enables transfer learning of semantics for a variety of natural language processing tasks. Word embeddings are typic…
Task-specific scores are often used to optimize for and evaluate the performance of conditional text generation systems. However, such scores are non-differentiable and cannot be used in the standard supervised learning paradigm. Hence, policy gradient methods are used since the gradient can be computed without requiri…
Training generative adversarial networks (GANs) on high quality (HQ) images involves important computing resources. This requirement represents a bottleneck for the development of applications of GANs. We propose a transfer learning technique for GANs that significantly reduces training time. Our approach consists of f…
MetFA aligns source and target domains for cross-device image classification.
problem Learning discriminative class boundaries across different domains.
method Distance metric guided feature alignment (MetFA) for domain-invariant and discriminative feature extraction.
result MetFA outperforms state-of-the-art methods in cross-device image classification.
This paper refines MMD for domain adaptation by balancing intra-class and inter-class distances.
problem Balancing intra-class and inter-class distances for better feature discriminability in domain adaptation.
method The paper theoretically proves two facts about MMD and proposes a novel discriminative MMD method to balance intra-class and inter-class distances.
result The proposed method improves feature discriminability and outperforms state-of-the-art methods.
We introduce an asymmetric distance in the space of learning tasks, and a framework to compute their complexity. These concepts are foundational for the practice of transfer learning, whereby a parametric model is pre-trained for a task, and then fine-tuned for another. The framework we develop is non-asymptotic, captu…