Beneficial perturbation network for continual learning

Sequential learning of multiple tasks in artificial neural networks using gradient descent leads to catastrophic forgetting, whereby previously learned knowledge is erased during learning of new, disjoint knowledge. Here, we propose a fundamentally new type of method - Beneficial Perturbation Network (BPN). We add task-dependent memory (biasing) units to allow the network to operate in different regimes for different tasks. We compute the most beneficial directions to train these units, in a manner inspired by recent work on adversarial examples. At test time, beneficial perturbations for a given task bias the network toward that task to overcome catastrophic forgetting. BPN is not only more parameter-efficient than network expansion methods, but also does not need to store any data from previous tasks, in contrast with episodic memory methods. Experiments on variants of the MNIST, CIFAR-10, CIFAR-100 datasets demonstrate strong performance of BPN when compared to the state-of-the-art.

[1]  Stefan Wermter,et al.  Continual Lifelong Learning with Neural Networks: A Review , 2018, Neural Networks.

[2]  Edith Lesburguères,et al.  Early Tagging of Cortical Networks Is Required for the Formation of Enduring Associative Memory , 2011, Science.

[3]  L. Squire,et al.  Retrograde amnesia and memory consolidation: a neurobiological perspective , 1995, Current Opinion in Neurobiology.

[4]  Matthew B. Blaschko,et al.  Encoder Based Lifelong Learning , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[5]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[6]  Sung Ju Hwang,et al.  Lifelong Learning with Dynamically Expandable Networks , 2017, ICLR.

[7]  Marc'Aurelio Ranzato,et al.  Gradient Episodic Memory for Continual Learning , 2017, NIPS.

[8]  Byoung-Tak Zhang,et al.  Overcoming Catastrophic Forgetting by Incremental Moment Matching , 2017, NIPS.

[9]  Dan Boneh,et al.  The Space of Transferable Adversarial Examples , 2017, ArXiv.

[10]  Laurent Itti,et al.  Closed-Loop Memory GAN for Continual Learning , 2018, IJCAI.

[11]  P. Frankland,et al.  The organization of recent and remote memories , 2005, Nature Reviews Neuroscience.

[12]  Moser Edvard,et al.  Pattern Separation in the Dentate Gyrus , 2009 .

[13]  Alex Krizhevsky,et al.  Learning Multiple Layers of Features from Tiny Images , 2009 .

[14]  R. French.  Catastrophic Forgetting in Connectionist Networks , 2006 .

[15]  Derek Hoiem,et al.  Learning without Forgetting , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[16]  Joan Bruna,et al.  Intriguing properties of neural networks , 2013, ICLR.

[17]  Razvan Pascanu,et al.  Overcoming catastrophic forgetting in neural networks , 2016, Proceedings of the National Academy of Sciences.

[18]  Michael McCloskey,et al.  Catastrophic Interference in Connectionist Networks: The Sequential Learning Problem , 1989 .

[19]  Laurent Itti,et al.  Closed-Loop GAN for continual Learning , 2018, IJCAI.

[20]  Ronald Kemker,et al.  FearNet: Brain-Inspired Model for Incremental Learning , 2017, ICLR.

[21]  Christoph H. Lampert,et al.  iCaRL: Incremental Classifier and Representation Learning , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).