Neural Architecture Retrieval

Authors: Xiaohuan Pei, Yanxi Li, Minjing Dong, Chang Xu

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive evaluations on both human-designed and synthesized neural architectures demonstrate the superiority of our algorithm. Such a dataset which contains 12k real-world network architectures, as well as their embedding, is built for neural architecture retrieval. Our project is available at www.terrypei.com/nnretrieval.
Researcher Affiliation Academia Xiaohuan Pei Department of Computer Science The University of Sydney, Australia xpei8318@uni.sydney.edu.au Yanxi Li Department of Computer Science The University of Sydney, Australia yanli0722@uni.sydney.edu.au Minjing Dong Department of Computer Science City University of Hong Kong, China minjdong@cityu.edu.hk Chang Xu Department of Computer Science The University of Sydney, Australia c.xu@sydney.edu.au
Pseudocode Yes A.2 ALGORITHM Algorithm 1 NAR Pre-training
Open Source Code No The paper states: "Our project is available at www.terrypei.com/nnretrieval." However, this is a general project website and not explicitly a source code repository or a statement that code is provided in supplementary materials, based on the prompt's criteria.
Open Datasets Yes For real-world neural architectures, we build a dataset with 12k different architectures collected from Hugging Face and Py Torch Hub, where each architecture is associated with an embedding for relevance computation.
Dataset Splits No We divided the pre-processed data records to train/test splits (0.9/0.1) stratified based on the fine-grained classes for testing the retrieval performance on the real-world neural architectures.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions general software components like GCNs, GAT, and specific frameworks like PyTorch and Hugging Face in the context of data collection, but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes In order to ensure the fairness of the implementation, we set the same hyperparameter training recipe for the baselines, and configure the same input channel, output channel, and number layers for the pre-training models. This encapsulation and modularity ensures that all differences in their retrieval performance come from a few lines of change. ... Table 5: Pre-training Recipes. EPs: Epochs; BS: Batch size; LR: Learning rate.