Triple-Memory Networks: A Brain-Inspired Method for Continual Learning

By Liyuan Wang, Bo Lei, Qian Li, Hang Su, Jun Zhu, Yi Zhong. 2020

64 citations recorded by the source corpus.

Continual acquisition of novel experience without interfering with previously learned knowledge, i.e., continual learning, is critical for artificial neural networks, while limited by catastrophic forgetting. A neural network adjusts its parameters when learning a new task but then fails to conduct the old tasks well. By contrast, the biological brain can effectively address catastrophic forgetting through consolidating memories as more specific or more generalized forms to complement each other, which is achieved in the interplay of the hippocampus and neocortex, mediated by the prefrontal cortex. Inspired by such a brain strategy, we propose a novel approach named triple-memory networks (TMNs) for continual learning. TMNs model the interplay of the three brain regions as a triple-network architecture of generative adversarial networks (GANs). The input information is encoded as specific representations of data distributions in a generator, or generalized knowledge of solving tasks in a discriminator and a classifier, with implementing appropriate brain-inspired algorithms to alleviate catastrophic forgetting in each module. Particularly, the generator replays generated data of the learned tasks to the discriminator and the classifier, both of which are implemented with a weight consolidation regularizer to complement the lost information in the generation process. TMNs achieve the state-of-the-art performance of generative memory replay on a variety of class-incremental learning benchmarks on MNIST, SVHN, CIFAR-10, and ImageNet-50.

Subfields: Artificial Intelligence, Cognitive Neuroscience, Meta & Continual Learning, Neuroscience

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

1. Han-Jia Ye
2. De-chuan Zhan
3. Da-Wei Zhou
4. Jun Zhu
5. Liyuan Wang
6. Hang Su
7. Ziwei Liu
8. Xingxing Zhang

In Cognitive Neuroscience

1. Qian Li
2. Liyuan Wang
3. Jun Zhu
4. Mingtian Zhang
5. Xingxing Zhang
6. Hang Su
7. Yi Zhong
8. Zsolt Kira

In Meta & Continual Learning

1. Liyuan Wang
2. Xingxing Zhang
3. Jun Zhu
4. Han-Jia Ye
5. De-chuan Zhan
6. Da-Wei Zhou
7. Hang Su
8. Qiwen Wang

In Neuroscience

1. Liyuan Wang
2. Jun Zhu
3. Qian Li
4. Zsolt Kira
5. De-chuan Zhan
6. Yen-Chang Hsu
7. Yi Zhong
8. Hang Su

Citations in the corpus

• A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, background): “Weight regularization [25], [55], [161], [162] and experience replay [25], [163] have been shown to be effective in mitigating catastrophic forgetting of generative models.”
• Incorporating neuro-inspired adaptability for continual learning in artificial intelligence (2023, background): “Representative strategies include selectively stabilizing parameters [7–11], recovering old data distributions [12–16] and allocating dedicated parameter subspaces [17, 18], etc.”
• SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model (2023, background): “Representative strategies include regularization-based approaches [19, 1, 51, 25, 41, 7], which preserve the old model and selectively stabilize changes of parameters or predictions; replaybased approaches [49, 30, 3, 44, 40, 39], which approximate and recover the previously-learned data distributio”
• AFEC: Active Forgetting of Negative Transfer in Continual Learning (2021, background): “Representative approaches include: weight regularization [12, 1, 36, 2], which selectively penalizes changes of the previously-learned parameters; parameter isolation [24, 10], which allocates a dedicated parameter subspace for each task; and memory replay [21, 9, 31, 32], which approximates and rec”
• Always Be Dreaming: A New Approach for Data-Free Class-Incremental Learning (2021, methodology): “An intuitive approach for DFCIL is to simultaneously train a generative model to be sampled for replay [27, 28, 51, 60].”
• Class-Incremental Learning: A Survey (2023, result): “The shade color in the last column denotes the subcategory, which is consistent with Figure 3.”