{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/multi-level-cnn-for-lung-nodule","title":"Multi-level CNN for lung nodule classification with Gaussian Process assisted hyperparameter optimization","arxiv_id":"1901.00276","date":"2019-01-02","proceeding":null,"authors":["Miao Zhang","Huiqi Li","Juan Lyu","Sai Ho Ling","Steven Su"],"abstract":"This paper investigates lung nodule classification by using deep neural\nnetworks (DNNs). Hyperparameter optimization in DNNs is a computationally\nexpensive problem, where evaluating a hyperparameter configuration may take\nseveral hours or even days. Bayesian optimization has been recently introduced\nfor the automatically searching of optimal hyperparameter configurations of\nDNNs. It applies probabilistic surrogate models to approximate the validation\nerror function of hyperparameter configurations, such as Gaussian processes,\nand reduce the computational complexity to a large extent. However, most\nexisting surrogate models adopt stationary covariance functions to measure the\ndifference between hyperparameter points based on spatial distance without\nconsidering its spatial locations. This distance-based assumption together with\nthe condition of constant smoothness throughout the whole hyperparameter search\nspace clearly violates the property that the points far away from optimal\npoints usually get similarly poor performance even though each two of them have\nhuge spatial distance between them. In this paper, a non-stationary kernel is\nproposed which allows the surrogate model to adapt to functions whose\nsmoothness varies with the spatial location of inputs, and a multi-level\nconvolutional neural network (ML-CNN) is built for lung nodule classification\nwhose hyperparameter configuration is optimized by using the proposed\nnon-stationary kernel based Gaussian surrogate model. Our algorithm searches\nthe surrogate for optimal setting via hyperparameter importance based\nevolutionary strategy, and the experiments demonstrate our algorithm\noutperforms manual tuning and well-established hyperparameter optimization\nmethods such as Random search, Gaussian processes with stationary kernels, and\nrecently proposed Hyperparameter Optimization via RBF and Dynamic coordinate\nsearch.","url_abs":"http://arxiv.org/abs/1901.00276v1","url_pdf":"http://arxiv.org/pdf/1901.00276v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"multi-level-cnn-for-lung-nodule","repo_url":"https://github.com/automl/fanova","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"bayesian-optimization","task_name":"Bayesian Optimization"},{"task_slug":"gaussian-processes","task_name":"Gaussian Processes"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"hyperparameter-optimization","task_name":"Hyperparameter Optimization"},{"task_slug":"lung-nodule-classification","task_name":"Lung Nodule Classification"}],"methods":[{"method_slug":"random-search","method_name":"Random Search"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}