Few-shot Generation via Recalling Brain-Inspired Episodic-Semantic Memory

Authors: Zhibin Duan, Zhiyi Lv, Chaojie Wang, Bo Chen, Bo An, Mingyuan Zhou

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments
Researcher Affiliation Academia National Laboratory of Radar Signal Processing Xidian University, Xi an, Shaanxi 710071, China xd_zhibin@163.com, xd_silly@163.com, bchen@mail.xidian.edu.cn Bo An Nanyang Technological University Mingyuan Zhou The University of Texas at Austin
Pseudocode Yes Algorithm 1 Semantic Memory recall and update algorithm
Open Source Code No The paper does not include any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes The experiments are conducted on five widely used benchmark datasets of various sizes, including binarized Omniglot [2], MNIST [35], DOUBLE-MNIST [36], Celeb A [37] and FS-CIFAR100 [38].
Dataset Splits No The split of training/testing set (also known as background-evaluation split) follows Lake et al. [2]... The paper mentions training and testing sets but does not explicitly provide details about a separate validation set split or its proportion.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments (e.g., GPU models, CPU types, or cloud computing instances).
Software Dependencies No The paper mentions various models and frameworks (e.g., VAE, Diffusion model, DDPM), but it does not provide specific version numbers for any software dependencies or libraries used in the implementation.
Experiment Setup Yes Model Settings: The details about model settings can be found in Appendix B. Memory size Across all datasets, we set the size of each semantic memory block (category) in the VSM-based model to 1, and each episodic memory block to 5. For models that exclusively depend on episodic memory (CNS + Episodic, SCHA + Episodic, and Episodic-diffusion), we establish a memory size of 10000, with the episodic memory functioning as a first-in, first-out queue.