Closed-Loop Memory GAN for Continual Learning

By Amanda Rios, Laurent Itti. 2018

64 citations recorded by the source corpus.

Sequential learning of tasks using gradient descent leads to an unremitting decline in the accuracy of tasks for which training data is no longer available, termed catastrophic forgetting. Generative models have been explored as a means to approximate the distribution of old tasks and bypass storage of real data. Here we propose a cumulative closed-loop memory replay GAN (CloGAN) provided with external regularization by a small memory unit selected for maximum sample diversity. We evaluate incremental class learning using a notoriously hard paradigm, single-headed learning, in which each task is a disjoint subset of classes in the overall dataset, and performance is evaluated on all previous classes. First, we show that when constructing a dynamic memory unit to preserve sample heterogeneity, model performance asymptotically approaches training on the full dataset. We then show that using a stochastic generator to continuously output fresh new images during training increases performance significantly further meanwhile generating quality images. We compare our approach to several baselines including fine-tuning by gradient descent (FGD), Elastic Weight Consolidation (EWC), Deep Generative Replay (DGR) and Memory Replay GAN (MeRGAN). Our method has very low long-term memory cost, the memory unit, as well as negligible intermediate memory storage.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models, Meta & Continual 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 Generative Models

In Meta & Continual Learning

Citations in the corpus

Replay in Deep Learning: Current Approaches and Missing Biological Elements (2021, neutral): “…learning, while the dynamic generative memory model (Ostapenko et al., 2019), the deep generative replay model (Shin et al., 2017), the Memory Replay GAN model (Wu et al., 2018), and the Closed-Loop GAN model (Rios & Itti, 2019) are all used to continually learn to generate images and scenes.”
Linear Mode Connectivity in Multitask and Continual Learning (2020, background): “Experience replay: Experience replay methods build and store a memory of the knowledge learned so far (Rebuffi et al., 2016; Lopez-Paz & Ranzato, 2017; Shin et al., 2017; Riemer et al., 2018; Rios & Itti, 2018; Zhang et al., 2019).”
GAN Memory with No Forgetting (2020, methodology): “Among these methods, generative replay is believed an effective and general strategy for lifelong learning problems [66, 88, 40, 78], as discussed above.”
Orthogonal Gradient Descent for Continual Learning (2019, background): “Shin et al. (2017); Kamra et al. (2017); Zhang et al. (2019); Rios and Itti (2018) learned a generative model to capture the data distribution of previous tasks, along with the current task’s data to train the new model so that the forgetting can be alleviated.”
Continual Learning for Robotics (2019, methodology): “Coresets have also been used to regularize the continual learning of a generative model in the CloGAN approach [130].”
Preserving Earlier Knowledge in Continual Learning with the Help of All Previous Feature Extractors (2021, background): “(GAN) [29], [30], [31], [32] to synthesize old data for each old class, such that plenty of synthetic but realistic data for each old class together with new classes of data can be used to train the new classifier.”
Continual Learning in Deep Networks: an Analysis of the Last Layer (2021, background): “Continual learning research field studied many strategies, such as rehearsal [26, 9, 2, 3, 52, 19, 6], generative replay [46, 53, 28, 30, 40], dynamic architecture [16, 42, 32] or regularization [22, 55, 41, 44].”