Unsupervised Meta-Learning through Latent-Space Interpolation in Generative Models

Authors: Siavash Khodadadeh, Sharare Zehtabian, Saeed Vahidian, Weijia Wang, Bill Lin, Ladislau Boloni

ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We tested the proposed algorithms on three few-shot learning benchmarks: (a) the 5-way Omniglot (Lake et al. (2011)), a benchmark for few-shot handwritten character recognition, (b) the 5-way Celeb A few-shot identity recognition, and (c) the Celeb A attributes dataset (Liu et al. (2015)) proposed as a few-shot learning benchmark by (Hsu et al. (2019)).
Researcher Affiliation Academia 1 Dept. of Computer Science University of Central Florida 2 Dept. of Electrical & Computer Engineering University of California San Diego
Pseudocode Yes Algorithm 1: LASIUM for unsupervised meta-learning task generation
Open Source Code Yes Our source code is also available on Github 2. Footnote 2: https://github.com/siavash-khodadadeh/Meta Learning-TF2.0
Open Datasets Yes We tested the proposed algorithms on three few-shot learning benchmarks: (a) the 5-way Omniglot (Lake et al. (2011)), a benchmark for few-shot handwritten character recognition, (b) the 5-way Celeb A few-shot identity recognition, and (c) the Celeb A attributes dataset (Liu et al. (2015)) proposed as a few-shot learning benchmark by (Hsu et al. (2019)).
Dataset Splits Yes During meta-learning, an additional set ,D(val) T , is attached to each task that contains another N K(val) data points separate from the ones in D(tr) T . We have exactly K(val) samples for each class in D(val) T as well. We set K(val) to be 15.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU/CPU models or memory specifications.
Software Dependencies No The paper describes neural network architectures and models used but does not provide specific software dependencies with version numbers (e.g., Python version, library versions like TensorFlow, PyTorch, or CUDA versions).
Experiment Setup Yes In this section, we report the hyperparameters of LASIUM-MAML in Table 5 and LASIUM-Proto Nets in Table 6 for Omniglot, Celeb A, Celeb A attributes and Mini-Image Net datasets. Table 5: LASIUM-MAML hyperparameters summary