{"id":1626,"date":"2024-07-02T12:30:52","date_gmt":"2024-07-02T12:30:52","guid":{"rendered":"https:\/\/2024.automl.cc\/?page_id=1626"},"modified":"2024-09-09T12:19:57","modified_gmt":"2024-09-09T12:19:57","slug":"recent-advances-in-meta-features-for-automated-single-objective-black-box-optimization","status":"publish","type":"page","link":"https:\/\/2024.automl.cc\/?page_id=1626","title":{"rendered":"Recent Advances in Meta-features for Automated Single-Objective Black-Box Optimization"},"content":{"rendered":"<div data-colibri-id=\"1626-c1\" class=\"style-783 style-local-1626-c1 position-relative\">\n  <!---->\n  <div data-colibri-component=\"section\" data-colibri-id=\"1626-c2\" id=\"custom\" class=\"h-section h-section-global-spacing d-flex align-items-lg-center align-items-md-center align-items-center style-788 style-local-1626-c2 position-relative\">\n    <!---->\n    <!---->\n    <div class=\"h-section-grid-container h-section-boxed-container\">\n      <!---->\n      <div data-colibri-id=\"1626-c3\" class=\"h-row-container gutters-row-lg-2 gutters-row-md-2 gutters-row-0 gutters-row-v-lg-2 gutters-row-v-md-2 gutters-row-v-2 style-791 style-local-1626-c3 position-relative\">\n        <!---->\n        <div class=\"h-row justify-content-lg-center justify-content-md-center justify-content-center align-items-lg-stretch align-items-md-stretch align-items-stretch gutters-col-lg-2 gutters-col-md-2 gutters-col-0 gutters-col-v-lg-2 gutters-col-v-md-2 gutters-col-v-2\">\n          <!---->\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-792-outer style-local-1626-c4-outer\">\n            <div data-colibri-id=\"1626-c4\" class=\"d-flex h-flex-basis h-column__inner h-px-lg-2 h-px-md-2 h-px-2 v-inner-lg-2 v-inner-md-2 v-inner-2 style-792 style-local-1626-c4 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100 align-self-lg-start align-self-md-start align-self-start\">\n                <!---->\n                <div data-colibri-id=\"1626-c5\" class=\"h-text h-text-component style-793 style-local-1626-c5 position-relative h-element\">\n                  <!---->\n                  <!---->\n                  <div class=\"\">\n                    <p>Date: 11.09.2024, 15:30-17:00<\/p>\n                    <p><span style=\"font-weight: 400; font-size: 16px; font-family: &quot;Open Sans&quot;; color: rgb(70, 112, 127);\">Room: 26-25\/105<\/span><\/p>\n                    <h2>Speakers<\/h2>\n                    <ul>\n                      <li>\n                        <a href=\"https:\/\/cs.ijs.si\/Gjorgjina_Cenikj\" style=\"font-family: &quot;Open Sans&quot;; font-weight: 400; font-size: 1em; color: rgb(3, 169, 244);\" class=\"customize-unpreviewable\">Gjorgjina Cenikj<\/a>\n                      <\/li>\n                      <li>\n                        <a href=\"https:\/\/cs.ijs.si\/Ana_Nikolikj\" style=\"font-family: &quot;Open Sans&quot;; font-weight: 400; font-size: 1em; color: rgb(3, 169, 244);\" class=\"customize-unpreviewable\">Ana Nikolikj<\/a>\n                      <\/li>\n                      <li>\n                        <a href=\"https:\/\/cs.ijs.si\/eftimov\" style=\"font-family: &quot;Open Sans&quot;; font-weight: 400; font-size: 1em; color: rgb(3, 169, 244);\" class=\"customize-unpreviewable\">Tome Eftimov<\/a>\n                      <\/li>\n                    <\/ul>\n                    <h2>Motivation<\/h2>\n                    <p>Algorithm selection aims to identify the best algorithm for a given problem instance, leveraging the strengths of different algorithms across various problems. However, selecting the optimal algorithm for an unseen instance is a complex\n                      challenge, attracting substantial research interest. This tutorial surveys key contributions to algorithm selection in single-objective continuous black-box optimization. It explores current efforts in using representation learning\n                      to derive meta-features for optimization problems, algorithms, and their interactions. Additionally, it examines the application of machine learning models for automated algorithm selection, configuration, and performance prediction.\n                      By analyzing these advancements, the tutorial highlights existing gaps in the field and proposes directions for enhancing meta-feature representations to improve algorithm selection efficacy.<\/p>\n                    <p>The tutorial will provide an overview of the recent trends in learning techniques for meta-features that can describe problem characteristics, algorithm characteristics, and algorithm-problem interactions (i.e., trajectory-based features).\n                      All of them will be presented in the single-objective black-box optimization scenario with opportunities to be transferred to other learning tasks. The problem landscape features are independent of any specific algorithm\u2019s behavior\n                      on a problem instance. Various methods have been developed to capture the characteristics of single-objective continuous optimization problems, categorized into high-level features, like those in Fitness Landscape Analysis (FLA),\n                      and low-level features, such as those in Exploratory Landscape Analysis (ELA), Topological Landscape Analysis (TLA), and deep learning-based approaches. High-level features are human-interpretable, while most low-level features are\n                      considered black-box features, which can enhance machine learning tasks. Algorithm features describe the characteristics of specific algorithm instances and can be derived from various sources, including source code, performance\n                      metrics, graph embeddings, and configuration settings. Additionally, trajectory-based features capture the interactions between the problem instance and the optimization algorithm by using samples generated during the algorithm\u2019s\n                      execution, providing a dynamic representation of this relationship. Machine learning models trained on these features can address algorithm selection tasks through multi-class classification, multi-label classification, regression,\n                      or ranking approaches. We will also provide a comprehensive comparison of machine learning techniques for algorithm selection using those features in single-objective optimisation.&nbsp;<\/p>\n                    <p>The tutorial will follow our survey article on meta-features for&nbsp;automated&nbsp;algorithm&nbsp;selection for&nbsp;black-box single objective&nbsp;continuous&nbsp;optimisation (\n                      <a href=\"https:\/\/arxiv.org\/abs\/2406.06629\" class=\"customize-unpreviewable\">https:\/\/arxiv.org\/abs\/2406.06629<\/a>).<\/p>\n                    <h2>Outline<\/h2>\n                    <ol>\n                      <li>Introduction to an automated algorithm selection (AAS) pipeline<\/li>\n                      <li>Calculating\/Learning problem landscape features<\/li>\n                      <li class=\"ql-indent-1\">Exploratory Landscape Analysis&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Topological Landscape Analysis&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Deep Learning-based features<\/li>\n                      <li class=\"ql-indent-2\">Feature used by Convolution Neural Networks&nbsp;<\/li>\n                      <li class=\"ql-indent-2\">Features learned using a point cloud transformer&nbsp;<\/li>\n                      <li class=\"ql-indent-2\">DoE2Vec features learned by using a variation autoencoder&nbsp;<\/li>\n                      <li class=\"ql-indent-2\">TransOpt features learned by using a transformer&nbsp;<\/li>\n                      <li class=\"ql-indent-2\">Deep-ELA features&nbsp;<\/li>\n                      <li>Calculating\/Learning algorithm features<\/li>\n                      <li class=\"ql-indent-1\">Algorithm Features Based on Source Code&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Algorithm Features Based on Performance&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Algorithm Features Based on Problem Landscape Features used by Performance Prediction Models&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Algorithm Features Using Graph Embeddings&nbsp;<\/li>\n                      <li>Learning trajectory-based features<\/li>\n                      <li class=\"ql-indent-1\">Trajectory-based features Based on Internal Algorithm Parameters&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Trajectory-based ELA features&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">DynamoRep features&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Opt2Vec features<\/li>\n                      <li class=\"ql-indent-1\">Iterative-based ELA features<\/li>\n                      <li class=\"ql-indent-1\">Local Optima Networks&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Search Trajectory Networks&nbsp;<\/li>\n                      <li class=\"ql-indent-1\">Probing trajectories&nbsp;<\/li>\n                      <li>Machine Learning techniques for AAS&nbsp;<\/li>\n                    <\/ol>\n                    <h2>Speakers<\/h2>\n                    <p>\n                      <br>\n                    <\/p>\n                  <\/div>\n                <\/div>\n              <\/div>\n            <\/div>\n          <\/div>\n        <\/div>\n      <\/div>\n    <\/div>\n  <\/div>\n  <div data-colibri-component=\"section\" data-colibri-id=\"1626-c6\" id=\"overlappable\" class=\"h-section h-section-global-spacing d-flex align-items-lg-center align-items-md-center align-items-center style-797 style-local-1626-c6 position-relative\">\n    <!---->\n    <!---->\n    <div class=\"h-section-grid-container h-section-boxed-container\">\n      <!---->\n      <div data-colibri-id=\"1626-c7\" class=\"h-row-container gutters-row-lg-0 gutters-row-md-0 gutters-row-0 gutters-row-v-lg-0 gutters-row-v-md-0 gutters-row-v-0 style-798 style-local-1626-c7 position-relative\">\n        <!---->\n        <div class=\"h-row justify-content-lg-center justify-content-md-center justify-content-center align-items-lg-stretch align-items-md-stretch align-items-stretch gutters-col-lg-0 gutters-col-md-0 gutters-col-0 gutters-col-v-lg-0 gutters-col-v-md-0 gutters-col-v-0\">\n          <!---->\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-799-outer style-local-1626-c8-outer\">\n            <div data-colibri-id=\"1626-c8\" class=\"d-flex h-flex-basis h-column__inner h-ui-empty-state-container h-px-lg-0 h-px-md-0 h-px-0 v-inner-lg-0 v-inner-md-0 v-inner-0 style-799 style-local-1626-c8 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100\">\n                <!---->\n              <\/div>\n            <\/div>\n          <\/div>\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-800-outer style-local-1626-c9-outer\">\n            <div data-colibri-id=\"1626-c9\" class=\"d-flex h-flex-basis h-column__inner h-px-lg-2 h-px-md-2 h-px-2 v-inner-lg-2 v-inner-md-2 v-inner-2 style-800 style-local-1626-c9 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100 align-self-lg-center align-self-md-center align-self-center\">\n                <!---->\n                <div data-colibri-id=\"1626-c10\" class=\"h-global-transition-all h-heading style-801 style-local-1626-c10 position-relative h-element\">\n                  <!---->\n                  <div class=\"h-heading__outer style-801 style-local-1626-c10\">\n                    <!---->\n                    <!---->\n                    <h4 class=\"\">Gjorgjina Cenikj<\/h4>\n                  <\/div>\n                <\/div>\n                <div data-colibri-id=\"1626-c11\" class=\"h-text h-text-component style-802 style-local-1626-c11 position-relative h-element\">\n                  <!---->\n                  <!---->\n                  <div class=\"\">\n                    <p>\n                      <a href=\"https:\/\/cs.ijs.si\/Gjorgjina_Cenikj\" style=\"font-family: &quot;Open Sans&quot;; font-weight: 400; font-size: 1em; color: rgb(3, 169, 244);\" class=\"customize-unpreviewable\">Gjorgjina Cenikj<\/a>&nbsp;is a young researcher at the Computer Systems Department at the Jo\u017eef Stefan Institute in Ljubljana, Slovenia. She is currently pursuing a PhD degree at the Jo\u017eef Stefan Postgraduate School, targeting the\n                      development of representation learning methodologies for single-objective numerical optimization problems and algorithms, with the goal of improving algorithm selection. She has been a major contributor to several feature-based characterizations\n                      of optimization problems Cenikj et al. (2024, 2023b,a); Petelin et al. (2022, 2024). Her research interests include machine learning, representation learning, natural language processing, and graph learning.<\/p>\n                  <\/div>\n                <\/div>\n              <\/div>\n            <\/div>\n          <\/div>\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-800-outer style-local-1626-c12-outer\">\n            <div data-colibri-id=\"1626-c12\" class=\"d-flex h-flex-basis h-column__inner h-px-lg-2 h-px-md-2 h-px-2 v-inner-lg-2 v-inner-md-2 v-inner-2 style-800 style-local-1626-c12 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100 align-self-lg-center align-self-md-center align-self-center\">\n                <!---->\n                <div data-colibri-id=\"1626-c13\" class=\"h-global-transition-all h-heading style-801 style-local-1626-c13 position-relative h-element\">\n                  <!---->\n                  <div class=\"h-heading__outer style-801 style-local-1626-c13\">\n                    <!---->\n                    <!---->\n                    <h4 class=\"\">Ana Nikolikj<\/h4>\n                  <\/div>\n                <\/div>\n                <div data-colibri-id=\"1626-c14\" class=\"h-text h-text-component style-802 style-local-1626-c14 position-relative h-element\">\n                  <!---->\n                  <!---->\n                  <div class=\"\">\n                    <p>\n                      <a href=\"https:\/\/cs.ijs.si\/Ana_Nikolikj\" style=\"font-family: &quot;Open Sans&quot;; font-weight: 400; font-size: 1em; color: rgb(3, 169, 244);\" class=\"customize-unpreviewable\">Ana Nikolikj<\/a>&nbsp;is a young researcher at the Computer Systems Department at the Jo\u017eef Stefan Institute in Ljubljana, Slovenia. She is working towards her PhD at the Jo\u017eef Stefan Postgraduate School, focusing on inventing methodologies\n                      to understand the behavior of single-objective numerical optimization algorithms via meta-learning. This is aimed at enhancing the process of algorithm performance prediction and algorithm selection. Her areas of interest encompass\n                      machine learning, representation learning, and methods for explainability. During her master thesis, she explored algorithm features based on explainable performance prediction models Nikolikj et al. (2022b,a).<\/p>\n                  <\/div>\n                <\/div>\n              <\/div>\n            <\/div>\n          <\/div>\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-805-outer style-local-1626-c15-outer\">\n            <div data-colibri-id=\"1626-c15\" class=\"d-flex h-flex-basis h-column__inner h-ui-empty-state-container h-px-lg-0 h-px-md-0 h-px-0 v-inner-lg-0 v-inner-md-0 v-inner-0 style-805 style-local-1626-c15 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100\">\n                <!---->\n              <\/div>\n            <\/div>\n          <\/div>\n        <\/div>\n      <\/div>\n    <\/div>\n  <\/div>\n  <div data-colibri-component=\"section\" data-colibri-id=\"1626-c16\" id=\"overlappable-2\" class=\"h-section h-section-global-spacing d-flex align-items-lg-center align-items-md-center align-items-center style-815 style-local-1626-c16 position-relative\">\n    <!---->\n    <!---->\n    <div class=\"h-section-grid-container h-section-boxed-container\">\n      <!---->\n      <div data-colibri-id=\"1626-c17\" class=\"h-row-container gutters-row-lg-0 gutters-row-md-0 gutters-row-0 gutters-row-v-lg-0 gutters-row-v-md-0 gutters-row-v-0 style-816 style-local-1626-c17 position-relative\">\n        <!---->\n        <div class=\"h-row justify-content-lg-center justify-content-md-center justify-content-center align-items-lg-stretch align-items-md-stretch align-items-stretch gutters-col-lg-0 gutters-col-md-0 gutters-col-0 gutters-col-v-lg-0 gutters-col-v-md-0 gutters-col-v-0\">\n          <!---->\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-817-outer style-local-1626-c18-outer\">\n            <div data-colibri-id=\"1626-c18\" class=\"d-flex h-flex-basis h-column__inner h-ui-empty-state-container h-px-lg-0 h-px-md-0 h-px-0 v-inner-lg-0 v-inner-md-0 v-inner-0 style-817 style-local-1626-c18 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100\">\n                <!---->\n              <\/div>\n            <\/div>\n          <\/div>\n          <div class=\"h-column h-column-container d-flex h-col-lg-auto h-col-md-auto h-col-auto style-818-outer style-local-1626-c19-outer\">\n            <div data-colibri-id=\"1626-c19\" class=\"d-flex h-flex-basis h-column__inner h-px-lg-2 h-px-md-2 h-px-2 v-inner-lg-2 v-inner-md-2 v-inner-2 style-818 style-local-1626-c19 position-relative\">\n              <!---->\n              <!---->\n              <div class=\"w-100 h-y-container h-column__content h-column__v-align flex-basis-100 align-self-lg-center align-self-md-center align-self-center\">\n                <!---->\n                <div data-colibri-id=\"1626-c20\" class=\"h-global-transition-all h-heading style-819 style-local-1626-c20 position-relative h-element\">\n                  <!---->\n                  <div class=\"h-heading__outer style-819 style-local-1626-c20\">\n                    <!---->\n                    <!---->\n                    <h4 class=\"\">Tome Eftimov<\/h4>\n                  <\/div>\n                <\/div>\n                <div data-colibri-id=\"1626-c21\" class=\"h-text h-text-component style-820 style-local-1626-c21 position-relative h-element\">\n                  <!---->\n                  <!---->\n                  <div class=\"\">\n                    <p>\n                      <a href=\"https:\/\/cs.ijs.si\/eftimov\" style=\"color: rgb(3, 169, 244); font-size: 1em; font-weight: 400; font-family: &quot;Open Sans&quot;;\">Tome Eftimov<\/a><span style=\"color: rgb(70, 112, 127); font-size: 16px; font-weight: 400; font-family: &quot;Open Sans&quot;;\">&nbsp;is a senior researcher at the Computer Systems Department at the Jo\u017eef Stefan Institute. He was a postdoctoral research fellow at Stanford University, USA, and a research associate at the University of California, San Francisco. His research interests include statistical data analysis, meta-learning, metaheuristics, natural language processing, representation learning, and machine learning. He has presented his work as 81 conference articles, 50 journal articles, and one Springer book published in 2022. He has been involved in courses on probability and statistics, and statistical data analysis. His work related to Deep Statistical Comparison was presented as a tutorial (i.e. IJCCI 2018, IEEE SSCI 2019, GECCO 2020, 2021, 2022, 2024, PPSN 2020, 2022, IEEE CEC 2021, 2022, 2023) or as an invited lecture to several international conferences and universities. Currently, he is supervising three Ph.D. students who are working on methodologies related to the proposed tutorial topic.<\/span><\/p>\n                  <\/div>\n                <\/div>\n              <\/div>\n            <\/div>\n          <\/div>\n        <\/div>\n      <\/div>\n    <\/div>\n  <\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Date: 11.09.2024, 15:30-17:00 Room: 26-25\/105 Speakers Gjorgjina Cenikj Ana Nikolikj Tome Eftimov Motivation Algorithm selection aims to identify the best algorithm for a given problem instance, leveraging the strengths of different algorithms across various problems. However, selecting the optimal algorithm for an unseen instance is a complex challenge, attracting substantial research interest. This tutorial surveys [&hellip;]<\/p>\n","protected":false},"author":9,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"page-templates\/full-width-page.php","meta":{"footnotes":""},"class_list":["post-1626","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/pages\/1626","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/2024.automl.cc\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1626"}],"version-history":[{"count":10,"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/pages\/1626\/revisions"}],"predecessor-version":[{"id":2598,"href":"https:\/\/2024.automl.cc\/index.php?rest_route=\/wp\/v2\/pages\/1626\/revisions\/2598"}],"wp:attachment":[{"href":"https:\/\/2024.automl.cc\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1626"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}