Architect, Regularize and Replay (ARR): a Flexible Hybrid Approach for
Continual Learning
- URL: http://arxiv.org/abs/2301.02464v1
- Date: Fri, 6 Jan 2023 11:22:59 GMT
- Title: Architect, Regularize and Replay (ARR): a Flexible Hybrid Approach for
Continual Learning
- Authors: Vincenzo Lomonaco, Lorenzo Pellegrini, Gabriele Graffieti, Davide
Maltoni
- Abstract summary: "Architect, Regularize and Replay" (ARR) is a hybrid generalization of the renowned AR1 algorithm and its variants.
It can achieve state-of-the-art results in classic scenarios (e.g. class-incremental learning) but also generalize to arbitrary data streams generated from real-world datasets such as CIFAR-100, CORe50 and ImageNet-1000.
- Score: 13.492896179777835
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: In recent years we have witnessed a renewed interest in machine learning
methodologies, especially for deep representation learning, that could overcome
basic i.i.d. assumptions and tackle non-stationary environments subject to
various distributional shifts or sample selection biases. Within this context,
several computational approaches based on architectural priors, regularizers
and replay policies have been proposed with different degrees of success
depending on the specific scenario in which they were developed and assessed.
However, designing comprehensive hybrid solutions that can flexibly and
generally be applied with tunable efficiency-effectiveness trade-offs still
seems a distant goal. In this paper, we propose "Architect, Regularize and
Replay" (ARR), an hybrid generalization of the renowned AR1 algorithm and its
variants, that can achieve state-of-the-art results in classic scenarios (e.g.
class-incremental learning) but also generalize to arbitrary data streams
generated from real-world datasets such as CIFAR-100, CORe50 and ImageNet-1000.
Related papers
- Dynamic Few-Shot Learning for Knowledge Graph Question Answering [3.116231004560997]
Large language models present opportunities for innovative Question Answering over Knowledge Graphs (KGQA)
To bridge this gap, solutions have been proposed that rely on fine-tuning or ad-hoc architectures, achieving good results but limited out-of-domain distribution generalization.
In this study, we introduce a novel approach called Dynamic Few-Shot Learning (DFL)
DFL integrates the efficiency of in-context learning and semantic similarity and provides a generally applicable solution for KGQA with state-of-the-art performance.
arXiv Detail & Related papers (2024-07-01T15:59:17Z) - Back to Basics: A Simple Recipe for Improving Out-of-Domain Retrieval in
Dense Encoders [63.28408887247742]
We study whether training procedures can be improved to yield better generalization capabilities in the resulting models.
We recommend a simple recipe for training dense encoders: Train on MSMARCO with parameter-efficient methods, such as LoRA, and opt for using in-batch negatives unless given well-constructed hard negatives.
arXiv Detail & Related papers (2023-11-16T10:42:58Z) - Domain-Agnostic Neural Architecture for Class Incremental Continual
Learning in Document Processing Platform [3.630365560970225]
Recent methods with learning gradient have been shown to struggle in such setups or have limitations like memory buffers.
We present a fully differentiable architecture based on the Mixture of Experts model, that enables the training of high-performance classifiers when examples from each class are presented separately.
We conducted exhaustive experiments that proved its applicability in various domains and ability to learn online in production environments.
arXiv Detail & Related papers (2023-07-11T16:01:44Z) - Massively Scalable Inverse Reinforcement Learning in Google Maps [3.1244966374281544]
Inverse reinforcement learning offers a powerful and general framework for learning humans' latent preferences in route recommendation.
No approach has successfully addressed planetary-scale problems with hundreds of millions of states and demonstration trajectories.
We revisit classic IRL methods in the routing context, and make the key observation that there exists a trade-off between the use of cheap, deterministic planners and expensive yet robust policies.
This insight is leveraged in Receding Horizon Inverse Planning (RHIP), a new generalization of classic IRL algorithms that provides fine-grained control over performance trade-offs via its planning horizon.
arXiv Detail & Related papers (2023-05-18T20:14:28Z) - Continual Predictive Learning from Videos [100.27176974654559]
We study a new continual learning problem in the context of video prediction.
We propose the continual predictive learning (CPL) approach, which learns a mixture world model via predictive experience replay.
We construct two new benchmarks based on RoboNet and KTH, in which different tasks correspond to different physical robotic environments or human actions.
arXiv Detail & Related papers (2022-04-12T08:32:26Z) - Rethinking Architecture Design for Tackling Data Heterogeneity in
Federated Learning [53.73083199055093]
We show that attention-based architectures (e.g., Transformers) are fairly robust to distribution shifts.
Our experiments show that replacing convolutional networks with Transformers can greatly reduce catastrophic forgetting of previous devices.
arXiv Detail & Related papers (2021-06-10T21:04:18Z) - LEADS: Learning Dynamical Systems that Generalize Across Environments [12.024388048406587]
We propose LEADS, a novel framework that leverages the commonalities and discrepancies among known environments to improve model generalization.
We show that this new setting can exploit knowledge extracted from environment-dependent data and improves generalization for both known and novel environments.
arXiv Detail & Related papers (2021-06-08T17:28:19Z) - Reinforcement Learning for Adaptive Mesh Refinement [63.7867809197671]
We propose a novel formulation of AMR as a Markov decision process and apply deep reinforcement learning to train refinement policies directly from simulation.
The model sizes of these policy architectures are independent of the mesh size and hence scale to arbitrarily large and complex simulations.
arXiv Detail & Related papers (2021-03-01T22:55:48Z) - Phase Retrieval using Expectation Consistent Signal Recovery Algorithm
based on Hypernetwork [73.94896986868146]
Phase retrieval is an important component in modern computational imaging systems.
Recent advances in deep learning have opened up a new possibility for robust and fast PR.
We develop a novel framework for deep unfolding to overcome the existing limitations.
arXiv Detail & Related papers (2021-01-12T08:36:23Z) - Learning to Localize in New Environments from Synthetic Training Data [26.194505911908585]
We present an approach that can generalize to new scenes by applying specific changes to the model architecture.
Our approach outperforms the 5-point algorithm using SIFT features on equally big images.
arXiv Detail & Related papers (2020-11-09T16:19:35Z) - Deep Keypoint-Based Camera Pose Estimation with Geometric Constraints [80.60538408386016]
Estimating relative camera poses from consecutive frames is a fundamental problem in visual odometry.
We propose an end-to-end trainable framework consisting of learnable modules for detection, feature extraction, matching and outlier rejection.
arXiv Detail & Related papers (2020-07-29T21:41:31Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.