FiT combines transfer and meta-learning for efficient few-shot image classification.
problem Few-shot image classification in personalized and federated learning settings.
method Combines transfer learning and meta-learning with fixed pretrained backbones and fine-tuned FiLM adapter layers.
result Achieves state-of-the-art accuracy on VTAB-1k benchmark with fewer than 1% of updateable parameters.
KM method reduces ConvNet parameters to 9% higher accuracy with minimal additional memory.
problem Expensive memory usage for training ConvNets on embedded devices.
method Kernel Modulation (KM) method that adapts all network parameters for each task.
result KM delivers up to 9% higher accuracy than other parameter-efficient methods.
Framework reuses pre-trained models for data-free transfer learning.
problem Challenges in retrieving source data for model training.
method Model Recycling Framework for parameter-efficient training.
result Makes multi-source data-free supervised transfer learning possible.
We introduce a novel method that enables parameter-efficient transfer and multi-task learning with deep neural networks. The basic approach is to learn a model patch - a small set of parameters - that will specialize to each task, instead of fine-tuning the last layer or the entire network. For instance, we show that l…
Model projection transfers convolutional network properties to feedforward networks.
problem Transferring properties between feedforward and convolutional networks.
method Unified node-level framework with tensor-valued activations, model projection.
result Projected CNN nodes inherit GFFN-style trainable structure.
Develops a method to ensure accuracy of few-shot transfer learning models.
problem Lack of generalization guarantees for low-data transfer learning.
method Trains a distribution over PEFT parameters using upstream tasks and samples plausible PEFTs for downstream tasks.
result Demonstrates non-vacuous generalization guarantees compared to existing methods in the low-shot regime.
Fine-tuning large pre-trained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transfer with adapter modules. Adapter modules yield a compact and extens…
FGSM is more stable in adversarially robust transfer learning than PGD.
problem Computational efficiency in adversarially robust transfer learning.
method Revisited use of FGSM in adversarial fine-tuning.
result FGSM is more stable and efficient in adversarial fine-tuning.
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.
A new framework for generative modeling using controlled vector fields.
problem Expressive modeling with limited parameters.
method Continuous-time modeling with modulated fixed vector fields and learned scalar controls.
result Expressive transport achieved with a small number of learned control channels.
Quaternion Conformer GAN (QC-GAN) is a parameter-efficient speech enhancement framework that combines a Quaternion Conformer generator with MetricGAN-based training.
problem Speech Enhancement
method Quaternion Conformer GAN
result Achieved a PESQ score of 3.48 with 0.89M parameters, comparable to state-of-the-art models at less than half their size.
TensorGuide improves LoRA efficiency and expressivity through joint tensor-train optimization.
problem Limited expressivity and generalization of standard LoRA.
method TensorGuide uses a unified tensor-train structure with controlled Gaussian noise to generate correlated low-rank matrices.
result TensorGuide achieves superior accuracy and scalability with fewer parameters compared to standard LoRA and TT-LoRA.
We introduce a new parameterization method for deep learning layers using spectral tensor train decomposition.
problem Efficiency and stability in deep learning models with weight matrix compression.
method Spectral Tensor Train Parameterization (STTP) of weight matrices.
result Improved compression and training stability in neural networks.
i-DenseNets improve parameter efficiency and performance in density estimation.
problem Improving parameter efficiency and performance in density estimation models.
method Invertible Dense Networks (i-DenseNets) with learnable weighted concatenation and Concatenated LipSwish activation function.
result i-DenseNets outperform Residual Flows and other flow-based models in bits per dimension.
New method improves reliability of depth estimation models.
problem Uncertainty quantification in large-scale vision models.
method Parameter-efficient Bayesian neural networks with PEFT methods.
result Combining PEFT methods with Bayesian inference enhances predictive performance.
Research reveals deep networks often learn low-rank structures, leading to more efficient training and fine-tuning.
problem Efficient training and deployment of large-scale deep learning models.
method Complementary theoretical perspectives on low-rank structures during training and convergence, and practical applications of LoRA and masked training.
result Understanding and exploiting low-rank structures can improve efficiency and effectiveness of training and fine-tuning.
Paper proposes CLAIR for efficient LLM fine-tuning across clients.
problem Fine-tuning large language models (LLMs) efficiently and collaboratively.
method Federated LoRA fine-tuning with Collaborative Low-rank Alignment and Identifiable Recovery (CLAIR).
result CLAIR achieves better performance and contamination detection compared to local fine-tuning.
New framework explains how larger pre-trained models reduce downstream learning sample complexity.
problem Understanding why larger pre-trained models reduce sample complexity in downstream tasks.
method Introducing a novel framework called Caulking inspired by PEFT methods.
result Improved pre-trained models provably decrease downstream task sample complexity.
We introduce a new function-preserving transformation for efficient neural architecture search. This network transformation allows reusing previously trained networks and existing successful architectures that improves sample efficiency. We aim to address the limitation of current network transformation operations that…
In this paper we explore the parameter efficiency of BERT arXiv:1810.04805 on version 2.0 of the Stanford Question Answering dataset (SQuAD2.0). We evaluate the parameter efficiency of BERT while freezing a varying number of final transformer layers as well as including the adapter layers proposed in arXiv:1902.00751. …
LaLoRA prevents forgetting in LoRA fine-tuning.
problem Catastrophic forgetting in fine-tuned models.
method LaLoRA applies Laplace approximation to LoRA weights for regularization.
result Improved learning-forgetting trade-off with controllable regularization strength.
Invertible DenseNets improve model efficiency and performance.
problem Improving model efficiency and performance in neural networks.
method Enforcing invertibility in DenseNets by satisfying the Lipschitz constraint and proposing a learnable concatenation.
result i-DenseNets outperform Residual Flows in negative log-likelihood on various datasets.
CAM-GAN improves GANs for continual learning with efficient feature map transformations.
problem Efficient continual learning for GANs with reduced parameter growth.
method Designing and leveraging parameter-efficient feature map transformations, including global and task-specific parameters, residual bias, and Fisher information matrix.
result Significantly improved model performance and high-quality samples with fewer parameters.
Transformers learn functionals from distributions without losing information.
problem Lack of rigorous mathematical theory supporting Transformer performance.
method Proposed a Transformer learning framework, attention operator, and distribution regression.
result Transformers can compress distributions into function representations without loss of information.
NeuroMAS treats multi-agent systems as neural networks for scalable, trainable coordination.
problem Designing multi-agent systems as hand-designed workflows is inefficient and inflexible.
method NeuroMAS treats multi-agent systems as a neural network architecture with reinforcement learning for scalable coordination.
result NeuroMAS improves significantly over multi-agent baselines and can be scaled progressively.
New hyperparameter ensembles boost neural network performance and uncertainty.
problem Improving neural network robustness and uncertainty quantification.
method Designing ensembles over both weights and hyperparameters, stratified across random initializations.
result Hyper-deep and hyper-batch ensembles outperform deep and batch ensembles on various architectures.
SPIRE enables efficient federated learning for diffusion models by separating client-specific embeddings from a shared backbone.
problem Large diffusion models are impractical for federated learning due to their size.
method SPIRE separates the network into a global backbone and client-specific embeddings, enabling efficient finetuning.
result SPIRE achieves parameter-efficient finetuning, updating only a small fraction of weights.
Improved neural network approximates analytic and L^p functions efficiently.
problem Efficiently approximating analytic and L^p functions using neural networks.
method Three-dimensional ReLU network architecture for sawtooth functions, improving approximation rates.
result Substantially improved exponential approximation rates for analytic functions and general L^p functions.
Paper analyzes transfer risk in transfer learning for finance.
problem Evaluate transferability of transfer learning in finance.
method Proposes transfer risk concept and applies to stock return prediction and portfolio optimization.
result Transfer risk correlates with transfer learning performance and identifies appropriate source tasks.
Transfer learning improves portfolio optimization by identifying transfer risk.
problem Financial portfolio optimization problem.
method Introduces transfer risk concept within transfer learning framework.
result Transfer risk is a significant indicator of transferability and enhances portfolio management efficiency.
CoLoRA leverages task similarity to boost fine-tuning efficiency.
problem Efficiently fine-tuning large foundation models with scarce labeled data.
method CoLoRA trains a shared adapter for task similarity and personalized adapters for user-specific tasks.
result CoLoRA significantly boosts fine-tuning performance when tasks are similar.
Mathematical framework for transfer learning feasibility and transfer risk.
problem Theoretical analysis of transfer learning.
method Reformulated transfer learning as an optimization problem, introduced transfer risk concept.
result Demonstrated the potential and benefits of incorporating transfer risk in transfer learning evaluation.
Kernelised flows improve density estimation and generation with fewer parameters.
problem Limited expressiveness of flow-based models due to invertibility constraints.
method Integrates kernels into normalising flows to enhance expressiveness and efficiency.
result Kernelised flows outperform neural network-based flows in parameter efficiency and low-data scenarios.
Transfer learning borrows knowledge from a source domain to facilitate learning in a target domain. Two primary issues to be addressed in transfer learning are what and how to transfer. For a pair of domains, adopting different transfer learning algorithms results in different knowledge transferred between them. To dis…
Paper fine-tunes LLMs for financial tasks using data fusion.
problem Improving LLMs for financial analysis tasks.
method Fine-tuned Llama3-8B and Mistral-7B using PEFT and LoRA, combined datasets for data fusion.
result Enhanced model performance across financial tasks.
LOFT separates subspace rotation and transformation for orthogonal fine-tuning.
problem Conflating subspace rotation and transformation in orthogonal fine-tuning.
method LOFT explicitly separates subspace rotation and transformation, using task-aware support selection.
result LOFT recovers principal-subspace orthogonal adaptation and improves efficiency-performance trade-off.
Pea-KD improves BERT student models by 4.4% on average in GLUE tasks.
problem Efficiently compressing BERT while maintaining performance.
method Parameter-efficient and accurate Knowledge Distillation (Pea-KD) with Shuffled Parameter Sharing (SPS) and Pretraining with Teacher's Predictions (PTP).
result Pea-KD improves student model's performance by 4.4% on average in four GLUE tasks.
Spatial Adapter adds structured spatial representation to frozen predictors.
problem Efficiently adding spatial structure to pre-trained models.
method Structured spatial decomposition and closed-form covariance for residual fields.
result Adapter improves spatial prediction and uncertainty quantification.
New research on limits of transfer learning, proving key selection and dependence requirements.
problem Insufficient theoretical foundation for transfer learning.
method Proved novel results on transfer learning, emphasizing selection of information and dependence between domains.
result Upper bound on improvement possible with transfer learning, highlighting the need for careful selection.
A new method to prune neural networks with iterative randomization improves efficiency.
problem Efficiency in pruning randomly initialized neural networks.
method Iteratively randomizing weight values to reduce parameter requirements.
result The method achieves remarkable performance with fewer parameters.
The paper analyzes phase transitions in transfer learning for perceptrons.
problem Understanding when transfer learning from a source task to a target task is beneficial.
method Theoretical analysis of a pair of related perceptron learning tasks.
result Reveals a phase transition from negative to positive transfer as task similarity changes.
Recent advances in semi-supervised learning with deep generative models have shown promise in generalizing from small labeled datasets (x,y) to large unlabeled ones (x). In the case where the codomain has known structure, a large unfeatured dataset (y) is potentially available.…
Transfer learning aims at improving the performance of target learners on target domains by transferring the knowledge contained in different but related source domains. In this way, the dependence on a large number of target domain data can be reduced for constructing target learners. Due to the wide application prosp…
This paper explores the connection between adversarial and knowledge transferability.
problem Understanding the factors affecting knowledge transferability.
method Theoretical analysis and practical metrics for adversarial transferability.
result Adversarial transferability and knowledge transferability are closely related.
Efficiently scales continuous kernels with sparse Fourier domain learning.
problem High computational and memory demands, spectral bias in continuous kernels.
method Sparse learning in the Fourier domain.
result Efficient scaling of continuous kernels, reduced computational and memory requirements, mitigated spectral bias.
Investigates transfer learning in spatial statistics.
problem Applying transfer learning to spatial statistics.
method Simple MLP models for spatial data.
result Potential of transfer learning in spatial statistics.
Proposes a transfer learning method for high-dimensional quantile regression.
problem Inadequate handling of heterogeneity and heavy tails in transfer learning.
method High-dimensional quantile regression framework with double transfer learning estimator.
result Established error bounds and valid confidence intervals for high-dimensional quantile regression coefficients.
TopoFisher learns topological summaries by maximizing Fisher information, improving parameter efficiency and inference quality.
problem Simulation-based inference misses key information in low-order statistics, especially for non-Gaussian fields.
method TopoFisher uses a differentiable persistent-homology pipeline that learns topological summaries by maximizing local Gaussian Fisher information.
result TopoFisher recovers much of the available information and outperforms fixed topological vectorizations in weak gravitational lensing.