Cognitively Inspired Learning of Incremental Drifting Concepts

Authors: Mohammad Rostami, Aram Galstyan

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our method on two sequential task learning settings: incremental learning and continual incremental learning.
Researcher Affiliation Academia Mohammad Rostami , Aram Galstyan University of Southern California {mrostami,galstyan}@isi.edu
Pseudocode Yes Algorithm 1 ICLA (λ, γ, τ)
Open Source Code Yes Our implementation is available as a supplement.
Open Datasets Yes We design two incremental learning experiments using the MNIST and the Fashion-MNIST datasets.
Dataset Splits No The paper mentions using "standard testing split" but does not explicitly provide specific percentages or counts for training, validation, or test splits. While these datasets have standard splits, the paper does not state them or cite how they were specifically used for reproduction.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running the experiments.
Software Dependencies No The paper does not provide specific software dependency details, such as library names with version numbers, needed to replicate the experiment.
Experiment Setup Yes Each task is learned in 100 epochs and at each epoch, the model performance is computed as the average classification rate over all the classes, observed before. We use a memory buffer with the fixed size of 100 for MB. We build an autoencoder by expanding a VGG-based classifier by mirroring the layers.