This page lists major projects on lifelong machine learning.


The Efficient Lifelong Learning (ELLA) Framework (Eric Eaton, et al., Univ. of Pennsylvania)

The ELLA framework provides a computationally efficient method for learning consecutive tasks. The approach uses techniques from sparse coding and online dictionary learning for online multi-task learning. It has a variety of theoretical guarantees on its performance and convergence, and has a computational complexity that is independent of the amount of data or number of tasks. The ELLA framework is a meta-learning approach, and can wrap around numerous base learners.  Multiple algorithms have been developed in this framework, including support for:


HORDE (Rich Sutton, et al., Univ. of Alberta)

HORDE is an architecture for lifelong learning from unsupervised sensorimotor data.  It consists of a large number of independent reinforcement learning sub-agents, or demons, each of which focuses on answering a single predictive or goal-oriented question about the environment.  HORDE trains all of the demons in parallel.  [AAMAS 2011]


The Never-Ending Language Learner (NELL) (Carnegie Mellon University)

NELL is part of the “Read the Web” project at CMU, which attempts to create a computer system that learns over time to read the web. Since January 2010,  NELL has been running continuously, extracting facts from text found on the web, and attempting to improve its reading competence to extract facts more accurately in the future.  It has accumulated an extensive database of facts from the web.  [Project Website]


L3ViSU: Lifelong Learning of Visual Scene Understanding (Christoph Lampert, et al., IST Austria)

The goal in the project is to develop and analyse algorithms that use continuous, open-ended machine learning from visual input data (images and videos) in order to interpret visual scenes on a level comparable to humans. [Project Website]