Learning Cumulatively to Become More Knowledgeable

By Geli Fei, Shuai Wang, B. Liu. 2016

103 citations recorded by the source corpus.

In classic supervised learning, a learning algorithm takes a fixed training data of several classes to build a classifier. In this paper, we propose to study a new problem, i.e., building a learning system that learns cumulatively. As time goes by, the system sees and learns more and more classes of data and becomes more and more knowledgeable. We believe that this is similar to human learning. We humans learn continuously, retaining the learned knowledge, identifying and learning new things, and updating the existing knowledge with new experiences. Over time, we cumulate more and more knowledge. A learning system should be able to do the same. As algorithmic learning matures, it is time to tackle this cumulative machine learning (or simply cumulative learning) problem, which is a kind of lifelong machine learning problem. It presents two major challenges. First, the system must be able to detect data from unseen classes in the test set. Classic supervised learning, however, assumes all classes in testing are known or seen at the training time. Second, the system needs to be able to selectively update its models whenever a new class of data arrives without re-training the whole system using the entire past and present training data. This paper proposes a novel approach and system to tackle these challenges. Experimental results on two datasets with learning from 2 classes to up to 100 classes show that the proposed approach is highly promising in terms of both classification accuracy and computational efficiency.

Subfields: Artificial Intelligence, 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 Meta & Continual Learning

Citations in the corpus

Learngene: From Open-World to Your Learning Task (2021, background): “Consequently, static and dedicated neural networks cannot resolve incrementally learning the unknown/unseen classes (Fei, Wang, and Liu 2016; Xu et al. 2019) without retraining from scratch and without catastrophic forgetting (French 1999), which is exactly what open-world learning (Bendale and…”
Incremental Learning from Scratch for Task-Oriented Dialogue Systems (2019, methodology): “The major challenges of the cumulative learning are finding unseen classes in the test set and updating itself efficiently to accommodate new concepts (Fei et al., 2016).”
DOC: Deep Open Classification of Text Documents (2017, methodology): “In our future work, we plan to improve the cumulative or incremental learning method in (Fei et al., 2016) to learn new classes without training on all past and new classes of data from scratch.”
Lifelong Generative Modeling (2017, methodology): “This supervised approach was later improved through the use of dynamic learning rates [115], core-sets [114] and multi-head classifiers [31].”
A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning (2020, background): “The recently introduced open world learning (Bendale and Boult, 2016) and the concurrently named cumulative learning (Fei et al., 2016) advance the pure open set identiﬁcation step by proposing to set aside the unknown unknowns and including them into a later active learning cycle.”
Open-world Learning and Application to Product Classification (2018, methodology): “It has been shown in [32] that DOC signicantly outperforms the methods CL-cbsSVM and cbsSVM in [9] and OpenMax in [3].”
Forward and Backward Knowledge Transfer for Sentiment Classification (2019, background): “Most related work to ours is lifelong learning (LL) (Mitchell et al., 2015; Ruvolo and Eaton, 2013; Chen et al., 2015; Fei et al., 2016; Isele et al., 2016; Xia et al., 2017; Shu et al., 2017; Xu et al., 2018; Sun et al., 2018).”
Unseen Class Discovery in Open-world Classification (2018, methodology): “After classification, (Bendale & Boult, 2015) and (Fei et al., 2016) also incrementally learn the unseen classes in the rejected examples after manually labeling.”
Towards Knowledgeable Supervised Lifelong Learning Systems (2020, methodology): “Learning about brain injury, for example, is limited by the training data.”
AI Autonomy: Self-initiated Open-world Continual Learning and Adaptation (2022, background): “…makes the closed-world assumption, which means that the classes of objects seen by the system in testing or application must have been seen during training (Fei, Wang, and Liu 2016; Bendale and Boult 2015; Liu 2020), i.e., there is nothing unexpected or novel occurring in testing or deployment.”
Open-World Continual Learning: Unifying Novelty Detection and Continual Learning (2023, background): “Open world learning has been studied by many researchers [54, 53, 6, 15, 37, 38].”
Self-Initiated Open World Learning for Autonomous AI Agents (2021, background): “Yet, some others have studied learning the novel objects or classes after they have been detected (Bendale and Boult 2015; Fei, Wang, and Liu 2016; Xu et al. 2019).”
HRN: A Holistic Approach to One Class Learning (2020, methodology): “Various forms of thresholding were used to detect anomalies [23, 24, 33, 51, 18, 76, 36, 20, 85].”
Lifelong machine learning: a paradigm for continuous learning (2017, background): “[7] extended this form of LML to cumulative learning, which, on encountering a new class, builds a new multi-class classifier that can classify all the past and the new classes.”
A Survey of Zero-Shot Learning (2019, background): “The cumulative learning [34] and class-incremental learning [123] methods have been proposed for problems in which labeled instances belonging to some previously unseen classes progressively appear after model learning.”