L2M

E17403

L2M is a DARPA research initiative focused on developing AI systems capable of continuous, lifelong learning and adaptation.

All labels observed (1)

Label Occurrences
L2M canonical 1

How this entity was disambiguated

Statements (44)

Predicate Object
instanceOf DARPA research program
artificial intelligence research initiative
abbreviationFor Lifelong Learning Machines program
surface form: Lifelong Learning Machines
applicationDomain autonomous vehicles
defense systems
intelligent agents
robotics
sensor processing
country United States of America
surface form: United States
focusesOn continual adaptation
continuous learning
lifelong learning in artificial intelligence
machine learning
robust AI systems
fullName Lifelong Learning Machines program
surface form: Lifelong Learning Machines
fundingAgency DARPA Information Innovation Office
goal develop AI systems capable of continuous, lifelong learning and adaptation
enable AI systems to learn from experience during deployment
enable AI to learn new tasks without catastrophic forgetting
improve robustness of AI in changing environments
reduce need for retraining from scratch
hasProperty aims to operate in dynamic, non-stationary environments
emphasizes learning during deployment
seeks biological inspiration from natural learning systems
targets efficient adaptation with limited data
targets robustness to unexpected situations
organizationTypeOfSponsor U.S. government research agency
relatedTo Lifelong Learning Machines program
surface form: DARPA Lifelong Learning Machines program

catastrophic forgetting mitigation techniques
continual learning research community
meta-learning
neuro-inspired computing
researchArea adaptive control
artificial intelligence
autonomous systems
continual learning
lifelong learning algorithms
machine learning
neuroscience-inspired AI
online learning
robust perception
transfer learning
sponsor Defense Advanced Research Projects Agency
surface form: DARPA

Defense Advanced Research Projects Agency

How these facts were elicited

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.