Fine-Grained Continual Learning

By Vincenzo Lomonaco, Davide Maltoni, Lorenzo Pellegrini. 2019

26 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Optimal Continual Learning has Perfect Memory and is NP-hard (2020, background): “As this implicit assumption is usually severely violated in practice, regularization-based CL algorithms often underperform, especially when the number of tasks is very large (van de Ven & Tolias, 2019; Lomonaco et al., 2019).”
Continual Reinforcement Learning in 3D Non-stationary Environments (2019, neutral): “More efficient consolidation techniques through regularization derived from EWC have been recently proposed [43, 51, 30, 27].”
Latent Replay for Real-Time Continual Learning (2019, methodology): “While iCaRL exhibits better performance than LWF and EWC (as reported in [15]), it is far from DSLDA, CWR* and AR1*.”
Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning (2020, methodology): “However, in many scenarios, it is unrealistic to deploy an agent without any world knowledge [42, 44], in part, since real-life non-i.”
Rehearsal-Free Continual Learning over Small Non-I.I.D. Batches (2019, methodology): “More details about the experiments and the hyperparameters used are provided in the additional materials of [19].”
RODEO: Replay for Online Object Detection (2020, neutral): “Continual learning (sometimes called incremental batch learning), is a much easier problem than streaming learning and has recently seen much success on classification and detection tasks [4, 6, 21, 25, 27, 35, 36, 41, 42, 51, 56].”
Online Continual Learning on Sequences (2020, methodology): “The saturation effect which characterizes the last training batches is due to the data distribution in NICv2 – 391 (see [44]): in particular, the lack of new instances for some classes (that already introduced all their data) slows down the accuracy trend and intensifies the effect of activations ag”
ContCap: A comprehensive framework for continual image captioning (2019, background): “This scenario can be referred to as class-incremental learning [16].”
Sequential Mastery of Multiple Visual Tasks: Networks Naturally Learn to Learn and Forget to Forget (2019, background): “[5] review existing visionbased benchmarks, including incremental training on MNIST [20], the CUB-200 dataset [21], and the CORe50 dataset [22, 23] for continual learning in an object recognition setting.”
Online Fast Adaptation and Knowledge Accumulation (OSAKA): a New Approach to Continual Learning (2020, background): “However, in many scenarios, it is unrealistic to deploy an agent without any world knowledge [43, 45], in part, since real-life non-i.”
ContCap: A scalable framework for continual image captioning (2019, background): “This scenario can be referred to as class-incremental learning [24].”
Generalized Class Incremental Learning (2020, unknown): “∗Equal contribution (1)[17, 18] refer to ‘phase’ as ‘batch’.”