Memory Aware Synapses: Learning what (not) to forget

By Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, Tinne Tuytelaars. 2017

2,096 citations recorded by the source corpus.

Humans can learn in a continuous manner. Old rarely utilized knowledge can be overwritten by new incoming information while important, frequently used knowledge is prevented from being erased. In artificial learning systems, lifelong learning so far has focused mainly on accumulating knowledge over tasks and overcoming catastrophic forgetting. In this paper, we argue that, given the limited model capacity and the unlimited new information to be learned, knowledge has to be preserved or erased selectively. Inspired by neuroplasticity, we propose a novel approach for lifelong learning, coined Memory Aware Synapses (MAS). It computes the importance of the parameters of a neural network in an unsupervised and online manner. Given a new sample which is fed to the network, MAS accumulates an importance measure for each parameter of the network, based on how sensitive the predicted output function is to a change in this parameter. When learning a new task, changes to important parameters can then be penalized, effectively preventing important knowledge related to previous tasks from being overwritten. Further, we show an interesting connection between a local version of our method and Hebb's rule,which is a model for the learning process in the brain. We test our method on a sequence of object recognition tasks and on the challenging problem of learning an embedding for predicting $ $ triplets. We show state-of-the-art performance and, for the first time, the ability to adapt the importance of the parameters based on unlabeled data towards what the network needs (not) to forget, which may vary depending on test conditions.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised 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 Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

The Curse of Recursion: Training on Generated Data Makes Models Forget (2023, methodology): “A typical way of preventing it is to use regularisations (Memory Aware Synpass [Aljundi et al., 2018]) or just rely on data ( e.g. Learning without Forgetting [Li and Hoiem, 2017]).”
Re-evaluating Continual Learning Scenarios: A Categorization and Case for Strong Baselines (2018, background): “) and use one of them in one task [17, 18, 19, 20].”
A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, background): “MAS [106] accumulates importance measures based on the sensitivity of predictive results to parameter changes, which is both online and unsupervised.”
Overcoming Multi-Model Forgetting (2019, positive): “An extreme case of sequential learning is lifelong learning, for which the solution to catastrophic forgetting developed by Aljundi et al. (2018) is also to prioritize the weight updates, with smaller updates for weights that are important for previously-learned tasks.”
Continuous Learning in Single-Incremental-Task Scenarios (2018, methodology): “Further ideas could be investigated in the future to quantify weight importance for old tasks such as exploiting the moving average of squared gradients already considered by methods like RMSprop [5] or Adam [11] or the Hebbian-like reinforcements between active neurons recently proposed in [1].”
Synaptic weight decay with selective consolidation enables fast learning without catastrophic forgetting (2019, background): “Forgetting can also be counterbalanced by reducing plasticity for weight conﬁgurations that are important in previous contexts, but assessing the importance of a synaptic weight requires additional information and memories (Kirkpatrick et al., 2017; Zenke et al., 2017; Aljundi et al., 2018).”
Continual Learning via Sequential Function-Space Variational Inference (2023, background): “Memory-aware synapses ( MAS ; Aljundi et al., 2018) use the gradient of the model output with respect to the parameters.”
Incorporating neuro-inspired adaptability for continual learning in artificial intelligence (2023, methodology): “Specifically, EWC [7], MAS [8] and SI [9] are synaptic regularization methods that selectively penalized parameter changes to preserve memory stability.”
The Ideal Continual Learner: An Agent That Never Forgets (2023, positive): “The point here is that, this classic line of research furnished the very rudimentary idea of what is recently known as selective forgetting or active forgetting in the deep continual learning context (Aljundi et al., 2018; Wang et al., 2021a; Shibata et al., 2021; Liu et al., 2022).”
SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model (2023, background): “Representative strategies include regularization-based approaches [19, 1, 50, 25, 40, 7], which preserve the old model and selectively stabilize changes of parameters or predictions; replay-based approaches [48, 29, 3, 43, 39, 38], which approximate and recover the previously-learned data…”
Beyond Not-Forgetting: Continual Learning with Backward Knowledge Transfer (2022, methodology): “, [12, 1, 18]) penalize the modification of important weights of old tasks; parameter-isolation based methods (e.”
On the Effectiveness of Lipschitz-Driven Rehearsal in Continual Learning (2022, methodology): “Among them, regularization methods work by introducing additional constraints in the form of loss terms; they are designed to limit the amount of total change either in parameter space [42, 89, 19, 2] or functional space [47, 10].”
Continual Learning by Modeling Intra-Class Variation (2022, background): “Hence, some recent work (Kirkpatrick et al., 2017; Aljundi et al., 2018; Zenke et al., 2017; Benzing, 2022) explores different measurements of the parameter importance.”
Class Gradient Projection For Continual Learning (2022, methodology): “Prevalent approaches to address the catastrophic forgetting problem in continual learning can be grouped into three broad categories: regularization-based approaches [2, 22, 30, 45], expansion-based approaches [17, 25, 31, 34, 36, 40], and memory-based approaches [3, 27, 33, 38].”
Long-Tailed Class Incremental Learning (2022, methodology): “R-EWC [22], Synaptic Intelligence (SI) [45], and Memory Aware Synapses (MAS) [3] adopt the same strategy but with different techniques to identify important weights.”
Meta-Learning with Less Forgetting on Large-Scale Non-Stationary Task Distributions (2022, neutral): “Many works mitigate catastrophic forgetting during the learning process [37,14,48,68,31,42,20,2,10,3,64].”
Anti-Retroactive Interference for Lifelong Learning (2022, background): “E-MAS-SDC proposed by [32] estimates the drift of previous tasks during the training of new tasks to make semantic drift compensation.”