Demystifying Randomly Initialized Networks for Evaluating Generative Models
Authors: Junghyuk Lee, Jun-Hyuk Kim, Jong-Seok Lee
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, we rigorously investigate the feature space of models with random weights in comparison to that of trained models. Furthermore, we provide an empirical evidence to choose networks for random features to obtain consistent and reliable results. Our results indicate that the features from random networks can evaluate generative models well similarly to those from trained networks, and furthermore, the two types of features can be used together in a complementary way. |
| Researcher Affiliation | Academia | Junghyuk Lee, Jun-Hyuk Kim, Jong-Seok Lee Yonsei University, South Korea {junghyuklee, junhyuk.kim, jong-seok.lee}@yonsei.ac.kr |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper. |
| Open Datasets | Yes | For the real-world image datasets, we use the Image Net (Deng et al. 2009) and FFHQ (Karras, Laine, and Aila 2019) datasets. We also use a non-natural cartoon image dataset that consists of Pok emon images. |
| Dataset Splits | No | The paper states that 50,000 generated images are used for evaluation but does not specify training, validation, or test dataset splits for any model being trained as part of this research, as the main focus is on evaluating existing generative models and randomly initialized networks (which are not trained in the typical sense for this study). |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment. |
| Experiment Setup | Yes | The Kaiming uniform (He et al. 2015) is used to initialize the weights. The evaluation results using the random features are averaged over five runs using different seeds for the random state initialization. |