APT-Gen generates tasks to help RL learn in hard problems.
problem Learning in hard exploration problems.
method APT-Gen uses a task generator to create tasks from a parameterized space, balancing performance and similarity to target tasks.
result APT-Gen outperforms baselines in grid world and robotic manipulation tasks.
The paper proposes a learning algorithm that improves adaptability and generalization.
problem Improving adaptability and generalization in learning models.
method Learning to meta-learn by meta-finetuning on related tasks before adapting to specific tasks.
result Learning to meta-learn improves adaptability and generalization across various tasks.
Multi-task learning is a learning paradigm which seeks to improve the generalization performance of a learning task with the help of some other related tasks. In this paper, we propose a regularization formulation for learning the relationships between tasks in multi-task learning. This formulation can be viewed as a n…
Meta-RL learns shared and task-specific information for quick adaptation.
problem Data inefficiency and limited generalization in deep RL.
method Task embedding and shared policy learned via SGD meta-learner.
result 3 to 4 times higher returns on novel tasks compared to baselines.
Small parameterized towers improve multi-task learning efficiency and generalization.
problem Balancing Pareto efficiency and generalization in multi-task learning.
method Under-parameterized self-auxiliaries for multi-task models.
result Small parameterized towers enhance Pareto efficiency in various multi-task applications.
Transformers can generalize to a large task family with only a few demonstrations.
problem Can learning from a small set of tasks generalize to a large task family?
method Investigating autoregressive compositional structure where each task is a composition of T T T operations, each from a finite family of D D D subtasks. result Transformers can generalize to D T D^T D T tasks with only O ~ ( D ) \widetilde{O}(D) O ( D ) demonstrations. Meta-World benchmark tests meta-reinforcement learning on 50 robotic tasks.
problem Current meta-reinforcement learning benchmarks are too narrow, limiting generalization.
method Proposes a new benchmark with 50 robotic manipulation tasks.
result State-of-the-art algorithms struggle with multiple tasks simultaneously.
Generalized Hindsight improves RL by reusing data from one task for another.
problem High sample complexity in reinforcement learning due to wasted uninformative data.
method Approximate inverse reinforcement learning to relabel behaviors with better-suited tasks.
result Efficient reuse of samples, reducing sample complexity on multi-task RL tasks.
Meta-learning improves by adding more data to tasks.
problem Overfitting to meta-training tasks impairs generalization to novel tasks.
method Task augmentation methods (MetaMix and Channel Shuffle) are proposed to improve generalization.
result MetaMix and Channel Shuffle outperform state-of-the-art results across multiple datasets.
Learning policies that generalize across multiple tasks is an important and challenging research topic in reinforcement learning and robotics. Training individual policies for every single potential task is often impractical, especially for continuous task variations, requiring more principled approaches to share and t…
Study MAML's generalization in varying tasks, proving bounds on error.
problem Bounding MAML's generalization error across tasks.
method Characterizes MAML's generalization error from two perspectives: recurring and unseen tasks.
result MAML's generalization error depends on the number of tasks and samples per task.
HGNN learns augmented features for deep multi-task learning.
problem Feature learning for deep multi-task learning.
method Hierarchical Graph Neural Network (HGNN) with two levels of graph neural networks.
result Significant performance improvement in classification tasks.
Improves shared encoder representations for better multi-task learning performance.
problem Improving quality of shared encoder representations in multi-task learning.
method Dummy Gradient norm Regularization (DGR) to decrease gradient norm of dummy task-specific predictors.
result DGR improves multi-task prediction performances and superior performance compared to existing methods.
Meta-learning shows negative transfer between tasks, which MetaCRL addresses.
problem Negative transfer between tasks in meta-learning.
method Structural Causal Models (SCMs) and MetaCRL to eliminate task confounders.
result MetaCRL achieves state-of-the-art performance in various benchmark datasets.
HML improves meta learning for diverse tasks.
problem Few-shot learning with heterogeneous tasks.
method Hierarchical factorization and dual-level training.
result HML outperforms existing methods in generalization.
CPVs enable efficient learning and generalization of complex tasks.
problem Learning and generalizing complex tasks efficiently.
method Compositional plan vectors (CPVs) for efficient task composition and generalization.
result CPVs enable efficient learning and generalization of complex tasks without additional supervision.
Analytic theory explains deep learning generalization and transfer learning.
problem Understanding how deep networks generalize and transfer knowledge across tasks.
method Developed an analytic theory for deep linear networks, predicting generalization and transfer.
result Deep networks learn task structure progressively, affecting generalization error.
Framework uses experience replay to prevent deep networks from forgetting past tasks.
problem Deep networks forget past tasks after learning new ones in sequential multitask learning.
method Generative model that couples current task with past learned tasks through a discriminative embedding space.
result Framework learns a shared abstract distribution across all tasks, preventing catastrophic forgetting.
Meta Omnium benchmarks few-shot learning across diverse vision tasks.
problem Evaluating generalization of few-shot learning across multiple vision tasks.
method Introduction of Meta Omnium dataset and evaluation of meta-learning algorithms.
result Meta-learning algorithms can generalize across diverse vision tasks.
Sharing information between multiple tasks enables algorithms to achieve good generalization performance even from small amounts of training data. However, in a realistic scenario of multi-task learning not all tasks are equally related to each other, hence it could be advantageous to transfer information only between …
New statistical guarantees for transfer learning with diverse tasks.
problem Statistical guarantees for transfer learning across different tasks.
method Formal analysis of t + 1 t+1 t + 1 tasks with shared representation. result Sample complexity reduction for new tasks using shared representation.
This paper reviews dMTL and methods for selecting auxiliary tasks.
problem Improving model performance for multiple tasks.
method Review of dMTL approaches and methods for selecting auxiliary tasks.
result Methods for selecting auxiliary tasks can improve dMTL performance.
New measure quantifies task difficulty for machine learning models.
problem Quantifying the inherent difficulty of machine learning tasks.
method Inductive bias complexity measure.
result Tasks requiring generalization over many dimensions are more difficult.
This work improves sample efficiency in meta-learning for nonlinear tasks.
problem Learning complex tasks efficiently with limited data.
method Subspace-based representations for nonlinear tasks.
result Subspace-based representations can be learned efficiently and improve future task performance.
The paper extends deep multi-task learning to diverse domains, finding shared functionality.
problem General problem solving from diverse deep learning tasks.
method Decomposes tasks into subproblems, optimizing sharing through stochastic algorithm.
result Joint learning across diverse domains and architectures improves performance.
Framework adapts to new tasks based on prior knowledge.
problem Models struggle to adapt to novel tasks without direct experience.
method Learned task representations and meta-mappings to transform them.
result Meta-mapping achieves 80-90% performance on novel tasks.
Unsupervised meta-learning speeds up reinforcement learning tasks.
problem Efficiently solving new reinforcement learning tasks.
method Formulating unsupervised meta-reinforcement learning and using mutual information for task proposals.
result Unsupervised meta-reinforcement learning effectively acquires accelerated procedures without manual task design.
Deep multi-task learning benefits from low intrinsic dimensionality, leading to better generalization.
problem Improving generalization in deep multi-task learning with high-dimensional models.
method Parametrizing multi-task networks in a low-dimensional space using random expansions and weight compression.
result First non-vacuous generalization bounds for deep multi-task networks are derived.
A method to learn multiple Atari games with a single agent.
problem Balancing multiple tasks in reinforcement learning.
method Adaptive task contribution to learning dynamics.
result Single agent surpasses human performance on 57 Atari games.
MGHRL learns to generate high-level meta strategies for new tasks.
problem Efficiency and generalization in meta-RL for wide task distributions.
method Generates high-level meta strategies over subgoals, leaving subtask learning independent.
result More efficient and generalized meta-learning from past experience.
MetaPerturb learns to improve generalization across different tasks and architectures.
problem Improving generalization on unseen data for diverse tasks and architectures.
method MetaPerturb is a meta-learned set-based perturbation function that improves generalization performance across heterogeneous tasks and architectures.
result MetaPerturb significantly outperforms baselines on most tasks and architectures with minimal increase in parameter size and no hyperparameters to tune.
Transformers learn to generalize out-of-distribution with diverse pretraining tasks.
problem Conditions for pretrained transformers to generalize out-of-distribution.
method Empirical study of task diversity and pretraining distribution.
result As task diversity increases, transformers transition from specialized to generalized solutions.
Study on neural networks' performance in sequential task learning.
problem Understanding the performance of neural networks in sequential task learning.
method Theoretical analysis of generalization performance in continual learning using statistical mechanical analysis of kernel ridge-less regression.
result Characteristic transitions from positive to negative transfer observed in neural networks.
Unified model for multiple healthcare tasks using graph structure.
problem Multiple healthcare tasks with shared graph structure.
method Graph-Convolutional Network (GCN) with multiple variational autoencoders.
result Boosted performance across all tasks, outperforming state-of-the-art.
Multi-task learning aims to learn multiple tasks jointly by exploiting their relatedness to improve the generalization performance for each task. Traditionally, to perform multi-task learning, one needs to centralize data from all the tasks to a single machine. However, in many real-world applications, data of differen…
Solves catastrophic forgetting in deep neural networks by generating task-relevant items.
problem Catastrophic forgetting in sequential learning of neural networks.
method Used a Generative Adversarial Network to generate items for rehearsing previous tasks.
result Deep network retains 1.67% absolute accuracy on CIFAR-10 and gains 0.24% on SVHN after learning multiple tasks.
Meta-learning framework uses task similarity through nonparametric kernel regression.
problem Limited tasks and outliers/dissimilar tasks hinder meta-learning performance.
method Nonparametric kernel regression to quantify and use task similarity.
result Meta-learning algorithm outperforms existing methods in task-limited settings.
Learned optimizers trained on many tasks improve model training efficiency and generalization.
problem Training models efficiently and effectively with minimal user intervention.
method Developed a neural network-based hierarchical optimizer trained on thousands of tasks.
result The learned optimizers generalize better to unseen tasks and exhibit unique behaviors.
Double descent in transfer learning explained for linear regression problems.
problem Understanding generalization errors in transferring parameters between overparameterized linear regression tasks.
method Analytical characterization of generalization error in terms of transfer learning factors.
result Generalization error follows a two-dimensional double descent trend controlled by transfer learning factors.
Study on meta-reinforcement learning generalization in high-dimensional tasks.
problem Generalization performance of meta-reinforcement learning algorithms in high-dimensional tasks.
method High-dimensional, procedurally generated environments.
result Meta-reinforcement learning algorithms exhibit strong overfitting on challenging tasks.
Infinite-Task Learning uses RKHSs to learn functions over hyperparameter space.
problem Learning a continuum of tasks with various loss functions.
method Utilizes operator-valued kernels and vector-valued RKHSs to control hyperparameters and constraints.
result Generalization guarantees and practical applications in classification, regression, and estimation.
Improves neural networks' ability to learn new tasks without forgetting earlier ones.
problem Preventing catastrophic forgetting in neural networks.
method Introducing a second discriminator in the GAN to generate important features for task retention.
result Significant reduction in catastrophic forgetting compared to standard methods.
This paper bounds meta-generalization gap using information theory.
problem Improving sample efficiency for new tasks in meta-learning.
method Information-theoretic upper bounds on meta-generalization gap for two meta-learning classes.
result Novel ITMI bounds for noisy iterative algorithms.
This thesis argues for learning representations from multiple similar tasks for better generalization.
problem Insufficient information in a single task for good generalization.
method Introduces the concept of an environment of a learner and uses a sample from it to learn a representation.
result Learning a representation from n n n tasks reduces the number of examples required for each task by a factor of n n n . This work evaluates task-agnostic exploration methods for fixed-batch learning.
problem Expensive real-world experience for robotics tasks.
method Fixed datasets for arbitrary task learning.
result Improved offline learning for robotics tasks.
Approach enables reinforcement learning agents to generalize from a single task to new tasks.
problem Catastrophic forgetting in reinforcement learning when facing a new task.
method Adversarial objective and semisupervised learning.
result The approach outperforms common baselines on Atari games.
New method learns auxiliary labels automatically for improved generalisation.
problem Improving generalisation in supervised learning without additional data.
method Trains two neural networks: label-generation and multi-task networks.
result MAXL outperforms single-task learning on 7 image datasets.
Learn generic latent relational graphs for better transfer learning.
problem Lack of transferable structured graphical representations in deep transfer learning.
method Unsupervisedly learned latent relational graphs from unlabeled data.
result Improves performance on various downstream tasks.