https://popups.uliege.be/3041-539x/index.php?id=806 Index terms fr 0 https://popups.uliege.be/3041-539x/index.php?id=804 This approach presents a possible solution to the stability-plasticity dilemma in incremental neural networks with a local insertion criterion. The main advantages are I) the capability of life-long learning, i.e., learning throughout the entire lifetime of a neural network, ii) stability in a stationary environment and iii) plasticity in a non-stationary environment, but only if the current knowledge does not fit the need of the task. Thus, the network structures its internal representation not like a copy of the environment but in order to fulfill the current task. Mon, 01 Jul 2024 13:58:45 +0200 Tue, 08 Oct 2024 14:33:18 +0200 https://popups.uliege.be/3041-539x/index.php?id=804 https://popups.uliege.be/3041-539x/index.php?id=3065 In incremental learning, it is necessary to conquer the dilemma of plasticity and stability. Because neural networks usually employ continuously distributed representation for state space, learning newly added data affects the existing memories. We apply a neural network with algebraic (lattice) structure to incremental learning, that has been proposed to model information processing in the dendrites of neurons. It has been proposed as a mathematical model of information processing in the dendrites of neurons. Because of the operation 'maximum' in lattice algebra weakening the continuously distributed representation, our proposed model succeeds in incremental learning. Fri, 06 Sep 2024 16:05:50 +0200 Fri, 06 Sep 2024 16:06:03 +0200 https://popups.uliege.be/3041-539x/index.php?id=3065