Lifelong Learning of Compositional Structures

By Jorge Armando Mendez Mendez, Eric Eaton. 2020

52 citations recorded by the source corpus.

A hallmark of human intelligence is the ability to construct self-contained chunks of knowledge and adequately reuse them in novel combinations for solving different yet structurally related problems. Learning such compositional structures has been a significant challenge for artificial systems, due to the combinatorial nature of the underlying search problem. To date, research into compositional learning has largely proceeded separately from work on lifelong or continual learning. We integrate these two lines of work to present a general-purpose framework for lifelong learning of compositional structures that can be used for solving a stream of related tasks. Our framework separates the learning process into two broad stages: learning how to best combine existing components in order to assimilate a novel problem, and learning how to adapt the set of existing components to accommodate the new problem. This separation explicitly handles the trade-off between the stability required to remember how to solve earlier tasks and the flexibility required to solve new tasks, as we show empirically in an extensive evaluation.

Subfields: Artificial Intelligence, Materials Chemistry, Representation & Self-Supervised Learning, Meta & Continual Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Materials Chemistry

In Representation & Self-Supervised Learning

In Meta & Continual Learning

Citations in the corpus

Task-Aware Information Routing from Common Representation Space in Lifelong Learning (2023, background): “On the other hand, Mendez & Eaton (2020) explicitly captured compositional structures in lifelong learning thereby enabling their reusability across tasks.”
Continual Learning via Local Module Composition (2021, methodology): “Our focus is on improving modular CL approaches, which despite their advantages, have only recently been studied in the CL literature [63, 92].”
Recent Advances of Continual Learning in Computer Vision: An Overview (2021, positive): “Mendez and Eaton [228] incorporated compositional learning into continual learning and proposed a generic framework that is agnostic to any continual learning algorithm.”
Optimizing Reusable Knowledge for Continual Learning via Metalearning (2021, background): “Others use different functions as components in the network, either Hypernetworks [39], Deep Artificial Neurons (DANs) [40], or Compositional Structures [41], so that the network components are more flexible during learning.”
Gradient-based Editing of Memory Examples for Online Task-free Continual Learning (2020, background): “NN) [36], Compositional Lifelong Learning [29], Graident Episodic Memory (GEM) [26] and Hindsight Anchor Learning (HAL) [7] respectively.”
A large-scale benchmark for few-shot program induction and synthesis (2021, methodology): “To improve this, more compositional methods to meta-learning are increasingly being proposed (Alet et al., 2018; Bengio et al., 2019; Ke et al., 2019; Mendez & Eaton, 2020; Ruis et al., 2020).”
Modular Lifelong Reinforcement Learning via Neural Composition (2022, methodology): “In the supervised setting, the agent is given a data set that is representative of the data distribution of the current task, which enables the agent to incorporate knowledge into the modules directly using the provided data after module selection (Mendez & Eaton, 2021).”
Sharing Less is More: Lifelong Learning in Deep Networks with Selective Layer Transfer (2020, background): “…of their high-level problems to this paper, these works on modular nets ensure the reusability of modules by the simultaneous access to data from multiple tasks in batch up front; only recently has the issue of modular reusability been explored in the lifelong setting (Mendez & Eaton, 2021).”
How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Composition (2022, methodology): “While the approach of Mendez and Eaton (2021) sidesteps forgetting in the structure over modules by making it task-specific, the controller of Qin et al. (2021) is susceptible to catastrophic forgetting.”
Continual Learning, Fast and Slow (2021, background): “SG-F [90] proposes a task-speciﬁc structural network that learns to update existing modules, combine modules, and add new modules.”
Continual Learning with Foundation Models: An Empirical Study of Latent Replay (2022, methodology): “…et al., 2017; Aljundi et al., 2017), dynamic architectures based methods (Yoon et al., 2017; Schwarz et al., 2018), which include modular methods (Mendez and EATON, 2021; Veniat et al., 2021; Ostapenko et al., 2021), and constrained optimization based methods (Lopez-Paz and Ranzato, 2017;…”