{"corpus_id":2377697,"paper_sha":"ad80ff20be876ad0fc4c041685519c284db5359e","doi":"10.1109/PRNI.2014.6858517","arxiv_id":"1407.5602","pmid":null,"pmcid":null,"mag_id":2056703339,"dblp_id":"conf/prni/DuboisHLPFFD14","acl_id":null,"title":"Predictive support recovery with TV-Elastic Net penalty and logistic regression: An application to structural MRI","year":2014,"publication_date":"2014-06-04","venue":"International Workshop on Pattern Recognition in NeuroImaging","journal":{"name":"2014 International Workshop on Pattern Recognition in Neuroimaging","pages":"1-4","volume":null},"journal_issn":null,"journal_title":null,"publication_types":["JournalArticle"],"pubmed_pub_types":null,"s2_fields_of_study":["Medicine","Computer Science","Mathematics"],"reference_count":12,"citation_count":18,"influential_citation_count":2,"is_open_access":true,"arxiv_categories":["stat.ML"],"arxiv_license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","arxiv_journal_ref":null,"mesh_headings":null,"chemicals":null,"comments_corrections":null,"source_flags":1,"s2_open_access_pdf_url":"http://arxiv.org/pdf/1407.5602","s2_open_access_landing_url":"https://www.semanticscholar.org/paper/ad80ff20be876ad0fc4c041685519c284db5359e","s2_open_access_license":null,"s2_open_access_status":"GREEN","pmc_open_access_pdf_url":null,"pmc_open_access_landing_url":null,"pmc_open_access_license":null,"pmc_open_access_status":null,"unpaywall_open_access_pdf_url":null,"unpaywall_open_access_landing_url":null,"unpaywall_open_access_license":null,"unpaywall_open_access_status":null,"abstract":"The use of machine-learning in neuroimaging offers new perspectives in early diagnosis and prognosis of brain diseases. Although such multivariate methods can capture complex relationships in the data, traditional approaches provide irregular (ℓ2 penalty) or scattered (ℓ1 penalty) predictive pattern with a very limited relevance. A penalty like Total Variation (TV) that exploits the natural 3D structure of the images can increase the spatial coherence of the weight map. However, TV penalization leads to non-smooth optimization problems that are hard to minimize. We propose an optimization framework that minimizes any combination of ℓ1, ℓ2, and TV penalties while preserving the exact ℓ1 penalty. This algorithm uses Nesterov's smoothing technique to approximate the TV penalty with a smooth function such that the loss and the penalties are minimized with an exact accelerated proximal gradient algorithm. We propose an original continuation algorithm that uses successively smaller values of the smoothing parameter to reach a prescribed precision while achieving the best possible convergence rate. This algorithm can be used with other losses or penalties. The algorithm is applied on a classification problem on the ADNI dataset. We observe that the TV penalty does not necessarily improve the prediction but provides a major breakthrough in terms of support recovery of the predictive brain regions.","claims":[{"public_id":"cl_d3e38796aa1266a2d1bf26450fd71d16","status":"active","text":"An optimization framework minimizes any combination of ℓ1, ℓ2, and TV penalties while preserving the exact ℓ1 penalty, using Nesterov's smoothing technique to approximate TV with a smooth function minimized by an exact accelerated proximal gradient algorithm.","confidence":0.95,"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_d3e38796aa1266a2d1bf26450fd71d16"},{"public_id":"cl_44beeed7b2d86c8602e15b286aa53cf4","status":"active","text":"An original continuation algorithm uses successively smaller values of the smoothing parameter to reach a prescribed precision while achieving the best possible convergence rate.","confidence":0.9,"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_44beeed7b2d86c8602e15b286aa53cf4"},{"public_id":"cl_7617d0756582927dfa91861528b246b7","status":"active","text":"TV penalty does not necessarily improve prediction but provides a major breakthrough in support recovery of predictive brain regions when applied to a classification problem on the ADNI dataset.","confidence":0.95,"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_7617d0756582927dfa91861528b246b7"}],"concepts":[{"public_id":"co_039ace408e15dbf7b309389e0d281d74","status":"active","name":"support recovery","description":"The identification of the true set of predictive brain regions from penalized regression models.","types":["evaluation criterion"],"aliases":["predictive support recovery"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_039ace408e15dbf7b309389e0d281d74"},{"public_id":"co_4d6fa7b7bf7d0a09d446f49bdd3d4317","status":"active","name":"continuation algorithm","description":"An original algorithm proposed here that applies successively smaller smoothing parameter values to reach a prescribed precision at the best convergence rate.","types":["optimization algorithm"],"aliases":[],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_4d6fa7b7bf7d0a09d446f49bdd3d4317"},{"public_id":"co_53db768ca9e52ea8e28b3cb6031d0b54","status":"active","name":"Nesterov's smoothing technique","description":"A method used here to approximate the non-smooth TV penalty with a smooth function, enabling accelerated gradient optimization.","types":["optimization technique"],"aliases":["Nesterov smoothing"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_53db768ca9e52ea8e28b3cb6031d0b54"},{"public_id":"co_599ec3323084650fdaf5f03b65842952","status":"active","name":"neuroimaging","description":"The field of brain imaging used here for early diagnosis and prognosis of brain diseases via machine learning methods.","types":["research field"],"aliases":[],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_599ec3323084650fdaf5f03b65842952"},{"public_id":"co_65502c897702d47f403a9086198fd8ee","status":"active","name":"ℓ1 penalty","description":"A sparsity-inducing regularization that produces scattered predictive patterns, preserved exactly in the proposed optimization framework.","types":["regularization method"],"aliases":["L1 penalty","lasso penalty"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_65502c897702d47f403a9086198fd8ee"},{"public_id":"co_7ebf62e1085e1983d10923d815a4014d","status":"active","name":"weight map","description":"A spatial map of regression coefficients across brain voxels whose coherence is shaped by the choice of regularization penalty.","types":["model output"],"aliases":["predictive pattern"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_7ebf62e1085e1983d10923d815a4014d"},{"public_id":"co_7fc091c330b75935e1d88d85a55527cb","status":"active","name":"Total Variation penalty","description":"A regularization penalty that exploits the natural 3D structure of images to increase spatial coherence of weight maps.","types":["regularization method"],"aliases":["TV penalty","TV"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_7fc091c330b75935e1d88d85a55527cb"},{"public_id":"co_7fd06c41cd8285e1400de4b17be82976","status":"active","name":"ℓ2 penalty","description":"A regularization that produces irregular predictive patterns, combined with TV and ℓ1 penalties in the proposed framework.","types":["regularization method"],"aliases":["L2 penalty","ridge penalty"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_7fd06c41cd8285e1400de4b17be82976"},{"public_id":"co_a15a4cd72c3f0a7b180f907ac481387e","status":"active","name":"ADNI dataset","description":"The Alzheimer's Disease Neuroimaging Initiative dataset used here to evaluate the proposed method on a brain disease classification problem.","types":["dataset"],"aliases":["Alzheimer's Disease Neuroimaging Initiative"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_a15a4cd72c3f0a7b180f907ac481387e"},{"public_id":"co_a497c89da7ab2f61658a7d791131483d","status":"active","name":"logistic regression","description":"A classification model used here with TV-Elastic Net penalization for brain disease classification.","types":["classification method"],"aliases":[],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_a497c89da7ab2f61658a7d791131483d"},{"public_id":"co_a892e7eeb7e9b5595ceef04bf7f0ea7d","status":"active","name":"structural MRI","description":"A neuroimaging modality providing 3D brain structure images used as input data in this paper's classification task.","types":["imaging modality"],"aliases":["sMRI"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_a892e7eeb7e9b5595ceef04bf7f0ea7d"},{"public_id":"co_b8a8cadfd02597ed7dba1dc9e348b9f9","status":"active","name":"accelerated proximal gradient algorithm","description":"An exact optimization algorithm used to minimize the combined loss and smoothed penalty after TV approximation.","types":["optimization algorithm"],"aliases":["proximal gradient method"],"contributors":[{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["extraction"],"url":"https://sah.borca.ai/u/322360f1c1"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["review"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_b8a8cadfd02597ed7dba1dc9e348b9f9"}],"external_ids":{"DOI":"10.1109/PRNI.2014.6858517","ArXiv":"1407.5602","PubMed":null,"PubMedCentral":null,"MAG":2056703339,"DBLP":"conf/prni/DuboisHLPFFD14","ACL":null},"open_access":{"is_open_access":true,"pdf_url":"http://arxiv.org/pdf/1407.5602","landing_url":"https://www.semanticscholar.org/paper/ad80ff20be876ad0fc4c041685519c284db5359e","source":"semantic_scholar","pdf_url_source":"semantic_scholar_open_access_pdf","license":null,"status":"GREEN","reason":null},"reference_availability":{"status":"available","references_indexed":true,"full_text_available":true,"full_text_source":"arxiv","count_basis":"semantic_scholar_metadata","extraction_status":"not_applicable","reason":null},"source":{"provider":"episteme2","base_corpus":"semantic_scholar_dump","freshness_mode":"unknown","basis":["semantic_scholar_metadata","postgres_metadata"],"limits":["paper metadata is based on indexed upstream scholarly datasets","claims and concepts are available only for extracted papers","absence of claims or concepts means no extracted graph data is available in this response"],"status":"available","degraded":false,"degraded_reasons":[],"diagnostics":{"status":"available","degraded":false,"degraded_reasons":[],"metadata_status":"available","graph_status":"available","abstract_status":"available"},"source_flags":1},"paper_id":630769,"paper_uid":"b314c01e-399e-4f38-afde-1f7bc5e5bc0c","canonical_identity":{"paper_id":630769,"paper_uid":"b314c01e-399e-4f38-afde-1f7bc5e5bc0c","identity_status":"available","lookup_basis":"semantic_scholar_external_id","compatibility_path":"corpus_id"},"url":"https://sah.borca.ai/papers/2377697"}