Learning to Learn with Gradients

By Chelsea Finn. 2018

110 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

A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms (2019, methodology): “meta-learning (Finn et al., 2017; Finn, 2018; Alet et al., 2018; Dasgupta et al., 2019), to Bayesian structure learning (Koller & Friedman, 2009; Heckerman et al.”
Meta-Learning Representations for Continual Learning (2019, methodology): “Improvements in automatic differentiation have made it more feasible to compute gradient-based meta-learning updates (Finn, 2018).”
Watch, Try, Learn: Meta-Learning from Demonstrations and Reward (2019, methodology): “Following the typical meta-learning problem statement [7], we will assume a distribution of tasks p(T ), from which the meta-training tasks {Ti} and held-out meta-test tasks {Tj} are drawn.”
Few-Shot Goal Inference for Visuomotor Learning and Planning (2018, methodology): “Following the meta-learning literature [47], we will consider the most general notion of a task, which can encapsulate a different objective, a different domain, a different environment, or a combination thereof.”
Deep Mixture of Experts via Shallow Embedding (2018, methodology): “aining the classiﬁcation network while our embedding objective is used to address the data scarcity issue of the weight generation network. Meta learning. Our work resides in the meta learning regime [9, 10, 36, 15] to learn the structure among previously seen tasks or concepts such that this learne”
A large-scale benchmark for few-shot program induction and synthesis (2021, background): “Moreover, deep learning-based meta-learning algorithms do not typically generalize broadly outside the data distribution, especially non-optimization-based approaches (Finn, 2018).”
Revisiting Meta-Learning as Supervised Learning (2020, methodology): “There are excellent overviews and surveys of metalearning (Thrun and Pratt, 1998; Vilalta and Drissi, 2002; Lemke et al., 2015; Vanschoren, 2018; Finn, 2018; Metz et al., 2018).”
When MAML Can Adapt Fast and How to Assist When It Cannot (2019, background): “ript is the line of work concerned with optimization-based meta-learning (OBML) algorithms in the few-shot regime, of which MAML is a particular instance. (Finn & Levine, 2018; Finn et al., 2017; Finn, 2018) While popular in practical applications, relatively few works have analysed the converge”
Meta-Learning with Implicit Gradients (2019, background): “and produce weight updates [25, 5, 33, 48] or predictions for new inputs [50, 12, 58, 40, 38], and optimization-based approaches that use bi-level optimization to embed learning procedures, such as gradient descent, into the meta-optimization problem [15, 13, 8, 60, 34, 17, 59, 23].”
NoRML: No-Reward Meta Learning (2019, unknown): “While MAML is successful at adapting policies to different tasks that are defined by reward changes (e.g. running forward vs. backward), it is less effective when adapting to other changes [9], such as dynamics changes, sensor drifts, or missing reward signals.”
Seven Defining Features of Terahertz (THz) Wireless Systems: A Fellowship of Communication and Sensing (2021, background): “, more expressive power means that the technqiue allows us to represent more sophisticated learning procedures [67].”
Meta-learning curiosity algorithms (2020, methodology): “Most work on meta-RL has focused on learning transferable feature representations or parameter values for quickly adapting to new tasks (Finn et al., 2017; Finn, 2018; Clavera et al., 2019) or improving performance on a single task (Xu et al., 2018; Veeriah et al., 2019).”
Self-Supervised Meta-Learning for Few-Shot Natural Language Classification Tasks (2020, methodology): “In contrast, we employ a ﬁne-tuning based meta-learning approach that enjoys nice properties like consistency which are important for good out-of-distribution generalization (Finn, 2018).”
How to Train Your MAML to Excel in Few-Shot Classification (2021, background): “Meta-learning, or learning to learn, is a sub-field of machine learning that attempts to search for the best learning strategy as the learning experiences increases [28, 49, 92, 95, 97].”