Using Task Features for Zero-Shot Knowledge Transfer in Lifelong Learning

By David Isele, Mohammad Rostami, Eric Eaton. 2016

93 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Lifelong Policy Gradient Learning of Factored Policies for Faster Training Without Forgetting (2020, background): “For PG-ELLA, we kept the same hyper-parameters as used for LPG-FTW, since they are used in exactly the same way for both methods.”
Efficient Lifelong Learning with A-GEM (2018, methodology): “The use of task descriptors for LLL has already been advocated by Isele et al. (2016) but using a sparse coding framework which is not obviously applicable to deep nets in a computationally efﬁ-cient way, and also by Lopez-Paz & Ranzato (2017) although they did not explore the use of compositional…”
Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learning (2017, background): “Isele et al. (2016) achieved zero-shot task generalization through dictionary learning with sparsity constraints.”
Forward and Backward Knowledge Transfer for Sentiment Classification (2019, background): “Most related work to ours is lifelong learning (LL) (Mitchell et al., 2015; Ruvolo and Eaton, 2013; Chen et al., 2015; Fei et al., 2016; Isele et al., 2016; Xia et al., 2017; Shu et al., 2017; Xu et al., 2018; Sun et al., 2018).”
A Survey of Zero-Shot Learning (2019, background): “In [63], it is combined with lifelong learning to learn new tasks with only descriptions.”
Active Lifelong Learning With "Watchdog" (2018, background): “Due to the fact that two-level task descriptor can improve knowledge transfer between multiple tasks in (Isele, Rostami, and Eaton 2016), we propose to learn a two-level dictionary called hierarchical dictionary in this paper.”
Transferring Autonomous Driving Knowledge on Simulated and Real Intersections (2017, background): “Recently it has been shown that deep reinforcement learning can improve over rule-based techniques (Isele et al., 2016), however it is unclear how well these techniques are able able to generalize to different scenarios and real systems.”
Sharing Less is More: Lifelong Learning in Deep Networks with Selective Layer Transfer (2020, background): “…assumes substantial similarity among tasks) or measuring conﬁguration frequency over re-lated tasks (which requires a notion of task similarity, such as via a task descriptor (Isele et al., 2016; Sinapov et al., 2015), or as determined dynamically by the task model’s relation to shared knowledge).”
Composable Planning with Attributes (2018, methodology): “Our approach in this work has a similar motivation to (Isele et al., 2016), where tasks are augmented with descriptors and featurized, and the coefﬁcients of the task features in a sparse dictionary are used to weight a set of vectors deﬁning the model for the associated task.”
Joint Dictionaries for Zero-Shot Learning (2017, methodology): “(Isele, Rostami, and Eaton 2016) used joint dictionary learning to learn a dynamical control system using high level task descriptors in an online lifelong zero-shot reinforcement learning setting.”
Learning Curriculum Policies for Reinforcement Learning (2018, background): “These approaches typically compute a similarity measure between the MDPs of the source and target task [2, 7], or learn a model of transferability that can be applied to novel source-target task pairs [11, 20].”
Learning Deep Parameterized Skills from Demonstration for Re-targetable Visuomotor Control (2019, methodology): “Other approaches achieve similar generalization through coupled dictionary learning [15], gating functions [20], and modular sub-networks [7].”
Using Task Descriptions in Lifelong Machine Learning for Improved Performance and Zero-Shot Transfer (2017, methodology): “In an earlier version of this work, we focused on RL tasks (Isele et al., 2016); here, we more fully explore the range of our approach, showing how it can be applied to regression, classiﬁcation, and RL problems.”
Zero-Shot Task Transfer (2019, background): “ial network. However, none of them operate in a zero-shot setting. We also found related work zero-shot task generalization in the context of reinforcement learning (RL) [29], or in lifelong learning [16]. An agent is validated based on its performance on unseen instructions or a longer instructions”
Learning Transferable Knowledge Through Embedding Spaces (2019, methodology): “Results of this chapter have been presented in [168; 61].”
How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Composition (2022, background): “, 2015), and another extension leverages task descriptors to achieve zero-shot transfer to unseen tasks (Isele et al., 2016; Rostami et al., 2020).”
Grounding Language for Transfer in Deep Reinforcement Learning (2017, background): “Isele, Rostami, and Eaton (2016) build on the ELLA framework (Ruvolo & Eaton, 2013; Ammar, Eaton, Ruvolo, & Taylor, 2014), and their key idea is to maintain two shared linear bases across tasks – one for the policy (L) and the other for task descriptors (D).”
State Abstractions for Lifelong Reinforcement Learning (2018, background): “Of growing interest in the RL literature is lifelong learning, in which an agent must interact with and solve many tasks over the course of a lifetime, as in Brunskill & Li (2014); Isele et al. (2016); Walsh et al. (2006) and Wilson et al. (2007).”