The Forget-me-not Process

By Kieran Milan, Joel Veness, James Kirkpatrick, Michael H. Bowling, Anna Koop, Demis Hassabis. 2016

22 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

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

In Probabilistic & Bayesian Methods

Citations in the corpus

Overcoming catastrophic forgetting in neural networks (2016, methodology): “It can be seen as an implementation of the flat forget me not algorithm described in [Kieran et al., 2016].”
Experience Replay for Continual Learning (2018, methodology): “…RL, Zenke et al. [32] develop a method similar to EWC that maintains estimates of the importance of weights for past tasks, Li and Hoiem [15] leverage a mixture of task-speciﬁc and shared parameters, and Milan et al. [19] develop a rigorous Bayesian approach for estimating unknown task boundaries.”
Policy Consolidation for Continual Reinforcement Learning (2019, methodology): “, 2017), for example, a method is employed for detecting these boundaries if they exist (Milan et al., 2016), it is still not practical for situations where the data distribution is changing continuously over time.”
Continual World: A Robotic Benchmark For Continual Reinforcement Learning (2021, background): “[34, 39]) to resolve this limitation, though we acknowledge the importance to extend the benchmark towards allowing and testing for task inference capabilities.”
Continual Unsupervised Representation Learning (2019, methodology): “In a similar fashion to existing techniques such as the Forget-Me-Not process (Milan et al., 2016) and Dirichlet process (Teh, 2010), we rely on a threshold to determine when to instantiate a new component.”
Life-Long Disentangled Representation Learning with Cross-Domain Latent Homologies (2018, background): “The majority of recent work in this direction has concentrated on implicit generative models [49, 48], or non-parametric approaches [35].”
Task Agnostic Continual Learning via Meta Learning (2019, methodology): “…the following baselines: None : the model is trained sequentially using Adam (Kingma & Ba, 2014) in the standard way without applying any continual learning method; Joint : all tasks seen so far are trained at the same time, this scheme usually sets the upper bound for continual learning methods.”