Task Agnostic Continual Learning Using Online Variational Bayes

Zeno, Chen; Golan, Itay; Hoffer, Elad; Soudry, Daniel

doi:10.48550/arxiv.1803.10123

Cited by 38 publications

(52 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To evaluate the proposed L2P, we closely follow the settings proposed in prior works [31,52,61], and conduct comprehensive experiments. In particular, we consider (1) the class-incremental setting, where the task identity is unknown during inference; (2) the domain-incremental setting, where the input domain shifts over time; (3) the taskagnostic setting, where there is no clear task boundary.…”

Section: Methodsmentioning

confidence: 99%

Learning to Prompt for Continual Learning

Wang¹,

Zhang²,

Lee³

et al. 2021

Preprint

View full text Add to dashboard Cite

The mainstream paradigm behind continual learning has been to adapt the model parameters to non-stationary data distributions, where catastrophic forgetting is the central challenge. Typical methods rely on a rehearsal buffer or known task identity at test time to retrieve learned knowledge and address forgetting, while this work presents a new paradigm for continual learning that aims to train a more succinct memory system without accessing task identity at test time. Our method learns to dynamically prompt (L2P) a pre-trained model to learn tasks sequentially under different task transitions. In our proposed framework, prompts are small learnable parameters, which are maintained in a memory space. The objective is to optimize prompts to instruct the model prediction and explicitly manage task-invariant and task-specific knowledge while maintaining model plasticity. We conduct comprehensive experiments under popular image classification benchmarks with different challenging continual learning settings, where L2P consistently outperforms prior state-ofthe-art methods. Surprisingly, L2P achieves competitive results against rehearsal-based methods even without a rehearsal buffer and is directly applicable to challenging taskagnostic continual learning. Source code is available at https://github.com/google-research/l2p.

show abstract

Section: Methodsmentioning

confidence: 99%

Learning to Prompt for Continual Learning

Wang¹,

Zhang²,

Lee³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Existing works in CL have proposed a variety of CL settings. In this section we explain some of the major CL settings that CLEAR adopts and refer readers to [1,22,44,53,59] for more thorough discussion of different variants of CL setups.…”

Section: Continual Learning Settingsmentioning

confidence: 99%

“…In this paper, we also adopt task-based sequential learning with a sequence of (same) 11-way classification tasks by splitting the temporal stream into 11 buckets, each consisting of a labeled subset for training and evaluation. However, it could be argued that in real-world, the model will not be informed about the task boundary (also called boundary-agnostic [33], task-free [1], or task-agnostic CL [59]). Such boundary-agnostic settings have been explored in recent works [1,4,23,59], in which a non-iid data stream continuously spits out new samples without a notion of task switch.…”

Section: Continual Learning Settingsmentioning

confidence: 99%

“…However, it could be argued that in real-world, the model will not be informed about the task boundary (also called boundary-agnostic [33], task-free [1], or task-agnostic CL [59]). Such boundary-agnostic settings have been explored in recent works [1,4,23,59], in which a non-iid data stream continuously spits out new samples without a notion of task switch. In this paper, we still assume a task-based sequential learning setting to ease benchmark design, but future works could adapt CLEAR to boundary-agnostic or task-free CL by processing data in an online streaming fashion using timestamps of CLEAR images.…”

Section: Continual Learning Settingsmentioning

confidence: 99%

See 1 more Smart Citation

The CLEAR Benchmark: Continual LEArning on Real-World Imagery

Lin

Song²,

Pathak

et al. 2022

Preprint

View full text Add to dashboard Cite

Continual learning (CL) is widely regarded as crucial challenge for lifelong AI. However, existing CL benchmarks, e.g. Permuted-MNIST and Split-CIFAR, make use of artificial temporal variation and do not align with or generalize to the realworld. In this paper, we introduce CLEAR, the first continual image classification benchmark dataset with a natural temporal evolution of visual concepts in the real world that spans a decade (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014). We build CLEAR from existing large-scale image collections (YFCC100M) through a novel and scalable low-cost approach to visio-linguistic dataset curation. Our pipeline makes use of pretrained vision-language models (e.g. CLIP) to interactively build labeled datasets, which are further validated with crowd-sourcing to remove errors and even inappropriate images (hidden in original YFCC100M). The major strength of CLEAR over prior CL benchmarks is the smooth temporal evolution of visual concepts with real-world imagery, including both high-quality labeled data along with abundant unlabeled samples per time period for continual semi-supervised learning. We find that a simple unsupervised pre-training step can already boost state-of-the-art CL algorithms that only utilize fully-supervised data. Our analysis also reveals that mainstream CL evaluation protocols that train and test on iid data artificially inflate performance of CL system. To address this, we propose novel "streaming" protocols for CL that always test on the (near) future. Interestingly, streaming protocols (a) can simplify dataset curation since today's testset can be repurposed for tomorrow's trainset and (b) can produce more generalizable models with more accurate estimates of performance since all labeled data from each time-period is used for both training and testing (unlike classic iid train-test splits).

show abstract

“…The method is however impractical. (Zeno et al, 2018) uses a logit masking related to ours but their context is based on the multi-head setting of continual learning and their goal is to activate only the head of which the samples within the new batch belong to. However, our approach is 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000…”

Section: Related Workmentioning

confidence: 99%

New Insights on Reducing Abrupt Representation Change in Online Continual Learning

Caccia¹,

Aljundi²,

Asadi³

et al. 2021

Preprint

View full text Add to dashboard Cite

We study the online continual learning paradigm, where agents must learn from a changing distribution with constrained memory and compute. Previous work often tackle catastrophic forgetting by overcoming changes in the space of model parameters. In this work we instead focus on the change in representations of previously observed data due to the introduction of previously unobserved class samples in the incoming data stream. We highlight the issues that arise in the practical setting where new classes must be distinguished between all previous classes. Starting from a popular approach, experience replay, we consider a metric learning based loss function, the triplet loss, which allows us to more explicitly constrain the behavior of representations. We hypothesize and empirically confirm that the selection of negatives used in the triplet loss plays a major role in the representation change, or drift, of previously observed data and can be greatly reduced by appropriate negative selection. Motivated by this we further introduce a simple adjustment to the standard cross entropy loss used in prior experience replay that achieves similar effect. Our approach greatly improves the performance of experience replay and obtains state-of-the-art on several existing benchmarks in online continual learning, while remaining efficient in both memory and compute.

show abstract

Task Agnostic Continual Learning Using Online Variational Bayes

Cited by 38 publications

References 15 publications

Learning to Prompt for Continual Learning

Learning to Prompt for Continual Learning

The CLEAR Benchmark: Continual LEArning on Real-World Imagery

New Insights on Reducing Abrupt Representation Change in Online Continual Learning

Contact Info

Product

Resources

About