Unified Probabilistic Deep Continual Learning through Generative Replay and Open Set Recognition

By Martin Mundt, Sagnik Majumder, Iuliia Pliushch, Visvanathan Ramesh. 2019

43 citations recorded by the source corpus.

Modern deep neural networks are well known to be brittle in the face of unknown data instances and recognition of the latter remains a challenge. Although it is inevitable for continual-learning systems to encounter such unseen concepts, the corresponding literature appears to nonetheless focus primarily on alleviating catastrophic interference with learned representations. In this work, we introduce a probabilistic approach that connects these perspectives based on variational inference in a single deep autoencoder model. Specifically, we propose to bound the approximate posterior by fitting regions of high density on the basis of correctly classified data points. These bounds are shown to serve a dual purpose: unseen unknown out-of-distribution data can be distinguished from already trained known tasks towards robust application. Simultaneously, to retain already acquired knowledge, a generative replay process can be narrowed to strictly in-distribution samples, in order to significantly alleviate catastrophic interference.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Ocean Engineering

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. Jiashi Feng
2. Li Yuan
3. Yujun Shi
4. Kristian Kersting
5. Quentin Delfosse
6. Steven Braun
7. Rui Gao
8. Weiwei Liu

In Computer Vision and Pattern Recognition

1. Jiashi Feng
2. Dacheng Tao
3. Luc Van Gool
4. Li Yuan
5. Lawrence Carin
6. Shuicheng Yan
7. Ming-Hsuan Yang
8. Yunpeng Chen

In Ocean Engineering

1. Rui Gao
2. Weiwei Liu
3. Martin Mundt
4. Kristian Kersting
5. Yunpeng Chen
6. Jiashi Feng

Citations in the corpus

• GAN Memory with No Forgetting (2020, background): “OCDVAE [54] unifies open-set recognition and VAE-based generative replay, whereas showing blurry generations for high-resolution images.”
• A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning (2020, methodology): “In our previous work (Mundt et al., 2022) we have shown how we can use equation 6 to reject samples from the prior z ∼ p(z) that do not fall into the obtained bounds of the aggregate posterior for generative rehearsal.”
• Continual Learning via Bit-Level Information Preserving (2021, methodology): “Replay Based Continual Learning These methods save representative examples of previous tasks in a replay buffer of limited size [19, 5, 13, 27, 37, 9] or train a generative model to generate samples of previous tasks [30, 31, 25, 32].”
• CLEVA-Compass: A Continual Learning EValuation Assessment Compass to Promote Research Transparency and Comparability (2021, methodology): “Correspondingly, this utility measure can either require presence of supervision, as shown in the class-specific meta-recognition model of OCDVAE, or can be entirely unsupervised.”
• DDGR: Continual Learning with Deep Diffusion-based Generative Replay (2023, methodology): “Classical GR methods use VAE or GAN as the generator for CL (Wu et al., 2018; Achille et al., 2018; Zhai et al., 2019; Wu et al., 2019; Ramapuram et al., 2020; Mundt et al., 2022).”