{ "access": { "embargo": { "active": false, "reason": null }, "files": "public", "record": "public", "status": "open" }, "created": "2021-08-05T17:29:58.811519+00:00", "custom_fields": { "journal:journal": { "title": "Neural Networks", "volume": "143" } }, "deletion_status": { "is_deleted": false, "status": "P" }, "files": { "count": 1, "enabled": true, "entries": { "2103.07492.pdf": { "checksum": "md5:e168aef7dad0e2f8f6d5d8711223b168", "ext": "pdf", "id": "a4f77d33-7775-4d8b-be7a-2acb22b7eb0e", "key": "2103.07492.pdf", "metadata": null, "mimetype": "application/pdf", "size": 1584864 } }, "order": [], "total_bytes": 1584864 }, "id": "5164245", "is_draft": false, "is_published": true, "links": { "access": "https://zenodo.org/api/records/5164245/access", "access_links": "https://zenodo.org/api/records/5164245/access/links", "access_request": "https://zenodo.org/api/records/5164245/access/request", "access_users": "https://zenodo.org/api/records/5164245/access/users", "archive": "https://zenodo.org/api/records/5164245/files-archive", "archive_media": "https://zenodo.org/api/records/5164245/media-files-archive", "communities": "https://zenodo.org/api/records/5164245/communities", "communities-suggestions": "https://zenodo.org/api/records/5164245/communities-suggestions", "doi": "https://doi.org/10.1016/j.neunet.2021.07.021", "draft": "https://zenodo.org/api/records/5164245/draft", "files": "https://zenodo.org/api/records/5164245/files", "latest": "https://zenodo.org/api/records/5164245/versions/latest", "latest_html": "https://zenodo.org/records/5164245/latest", "media_files": "https://zenodo.org/api/records/5164245/media-files", "parent": "https://zenodo.org/api/records/5164244", "parent_doi": "https://zenodo.org/doi/", "parent_html": "https://zenodo.org/records/5164244", "requests": "https://zenodo.org/api/records/5164245/requests", "reserve_doi": "https://zenodo.org/api/records/5164245/draft/pids/doi", "self": "https://zenodo.org/api/records/5164245", "self_doi": "https://zenodo.org/doi/10.1016/j.neunet.2021.07.021", "self_html": "https://zenodo.org/records/5164245", "self_iiif_manifest": "https://zenodo.org/api/iiif/record:5164245/manifest", "self_iiif_sequence": "https://zenodo.org/api/iiif/record:5164245/sequence/default", "versions": "https://zenodo.org/api/records/5164245/versions" }, "media_files": { "count": 0, "enabled": false, "entries": {}, "order": [], "total_bytes": 0 }, "metadata": { "creators": [ { "affiliations": [ { "name": "University of Pisa" } ], "person_or_org": { "family_name": "Andrea Cossu", "name": "Andrea Cossu", "type": "personal" } }, { "affiliations": [ { "name": "University of Pisa" } ], "person_or_org": { "family_name": "Antonio Carta", "name": "Antonio Carta", "type": "personal" } }, { "affiliations": [ { "name": "University of Pisa" } ], "person_or_org": { "family_name": "Vincenzo Lomonaco", "name": "Vincenzo Lomonaco", "type": "personal" } }, { "affiliations": [ { "name": "University of Pisa" } ], "person_or_org": { "family_name": "Davide Bacciu", "name": "Davide Bacciu", "type": "personal" } } ], "description": "
Learning continuously during all model lifetime is fundamental to deploy machine learning solutions robust to drifts in the data distribution. Advances in Continual Learning (CL) with recurrent neural networks could pave the way to a large number of applications where incoming data is non stationary, like natural language processing and robotics. However, the existing body of work on the topic is still fragmented, with approaches which are application-specific and whose assessment is based on heterogeneous learning protocols and datasets. In this paper, we organize the literature on CL for sequential data processing by providing a categorization of the contributions and a review of the benchmarks. We propose two new benchmarks for CL with sequential data based on existing datasets, whose characteristics resemble real-world applications.
\n\nWe also provide a broad empirical evaluation of CL and Recurrent Neural Networks in class-incremental scenario, by testing their ability to mitigate forgetting with a number of different strategies which are not specific to sequential data processing. Our results highlight the key role played by the sequence length and the importance of a clear specification of the CL scenario.
", "funding": [ { "award": { "acronym": "TEACHING", "id": "00k4n6c32::871385", "identifiers": [ { "identifier": "https://cordis.europa.eu/projects/871385", "scheme": "url" } ], "number": "871385", "program": "H2020", "title": { "en": "A computing toolkit for building efficient autonomous applications leveraging humanistic intelligence" } }, "funder": { "id": "00k4n6c32", "name": "European Commission" } } ], "languages": [ { "id": "eng", "title": { "en": "English" } } ], "publication_date": "2021-08-05", "publisher": "Zenodo", "resource_type": { "id": "publication-article", "title": { "de": "Zeitschriftenartikel", "en": "Journal article" } }, "rights": [ { "description": { "en": "The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited." }, "icon": "cc-by-icon", "id": "cc-by-4.0", "props": { "scheme": "spdx", "url": "https://creativecommons.org/licenses/by/4.0/legalcode" }, "title": { "en": "Creative Commons Attribution 4.0 International" } } ], "subjects": [ { "subject": "continual learning; recurrent neural networks" } ], "title": "Continual learning for recurrent neural networks: An empirical evaluation" }, "parent": { "access": { "owned_by": { "user": 246602 } }, "communities": { "default": "89ee3a47-3e75-4d1b-b596-2ce9e9ef22ab", "entries": [ { "access": { "member_policy": "open", "members_visibility": "public", "record_policy": "open", "review_policy": "open", "visibility": "public" }, "children": { "allow": false }, "created": "2021-03-10T14:25:18.372885+00:00", "custom_fields": {}, "deletion_status": { "is_deleted": false, "status": "P" }, "id": "89ee3a47-3e75-4d1b-b596-2ce9e9ef22ab", "links": {}, "metadata": { "curation_policy": "", "page": "TEACHING is a research project that designs a computing platform and the associated software toolkit supporting the development and deployment of autonomous, adaptive and dependable CPSoS applications, allowing them to exploit a sustainable human feedback to drive, optimize and personalize the provisioning of their services.
\r\n\r\nTEACHING receives funding from the European Union’s Horizon 2020 Research and Innovation program under grant agreement No 871385.
\r\n\r\nFor more information, please visit: https://www.teaching-h2020.eu/
", "title": "TEACHING - A computing toolkit for building efficient autonomous applications leveraging humanistic intelligence" }, "revision_id": 0, "slug": "teaching-h2020", "updated": "2021-03-11T08:04:33.199157+00:00" }, { "access": { "member_policy": "open", "members_visibility": "restricted", "record_policy": "open", "review_policy": "closed", "visibility": "public" }, "children": { "allow": true }, "created": "2022-11-23T15:53:29.436323+00:00", "custom_fields": {}, "deletion_status": { "is_deleted": false, "status": "P" }, "id": "f0a8b890-f97a-4eb2-9eac-8b8a712d3a6c", "links": {}, "metadata": { "curation_policy": "The EU Open Research Repository serves as a repository for research outputs (data, software, posters, presentations, publications, etc) which have been funded under an EU research funding programme such as Horizon Europe, Euratom or earlier Framework Programmes.
\nThe community is managed by CERN on behalf of the European Commission.
\nZenodo’s general policies and Terms of Use apply to all content.
\nThe EU Open Research Repository accepts all digital research objects which is a research output stemming from one of EU’s research and innovation funding programmes. The funding programmes currently include:
\nHorizon Europe (including ERC, MSCA), earlier Framework Programmes (eg Horizon 2020) as well as Euratom.
\nIn line with the principle as open as possible, as closed as necessary both public and restricted content is accepted. See note on how Zenodo handles restricted content.
\nEU programme beneficiaries are eligible to submit content to the community. The community supports three types of content submissions:
\nSubmission via an EU Project Community (through user interface or programmatic APIs).
\nSubmission directly to the EU Open Research Repository.
\nAutomated harvesting from existing Zenodo content.
\nA representative of an EU project may request an EU Project Community and invite other project participants as members of the community. The project community is linked to one or more European Commission grants. All records in the project community are automatically integrated into the EU Open Research Repository immediately upon acceptance into the project community.
\nAny user may submit a record directly to the EU Open Research Repository. The submission will be moderated by Zenodo staff for compliance with the minimal required metadata requirements and its correctness.
\nRecords found among Zenodo’s existing content will on a regular basis automatically be integrated if they are found to comply with the requirements. The submissions through this method are integrated into the EU Open Research Repository with delay in a fully automated way.
\nRecords in the EU Open Research Repository are required to comply with the following minimal metadata requirements:
\nVisibility: Both public and restricted (with or without embargo and/or access request)
\nResource types: All resource types.
\nLicenses: Public and embargoed records MUST specify a license.
\nFunding information: Records MUST specify at least one grant from the European Commission.
\nCreators: Creators SHOULD be identified with a persistent identifier (e.g. ORCID, GND, …), and affiliations SHOULD be identified with a persistent identifier (e.g. ROR, ISNI, …)
\nSubjects: Records SHOULD specify one or more fields of science from the European Science Vocabulary.
\nAll submissions will undergo automated curation checks for compliance with the policy. Submissions through project communities are reviewed by the project community. Submission directly to the EU Open Research Repository is reviewed by Zenodo staff.
\nCommunity curators may at any point edit metadata of the records in the community without notice through human or automated processing. The curators may at their sole discretion remove records from the community that are deemed not to comply with the content and curation policy or which are deemed of insufficient quality.
\nThe content and curation policy is subject to change by the community owner at any time and without notice, other than through updating this page.
", "description": "Open repository for EU-funded research outputs from Horizon Europe, Euratom and earlier Framework Programmes.", "organizations": [ { "id": "00k4n6c32" } ], "page": "The EU Open Research Repository is a Zenodo-community dedicated to fostering open science and enhancing the visibility and accessibility of research outputs funded by the European Union. The community is managed by CERN on behalf of the European Commission.
\nThe mission of the repository is to support the implementation of the EU's open science policy, providing a trusted and comprehensive space for researchers to share their research outputs such as data, software, reports, presentations, posters and more. The EU Open Research Repository simplifies the process of complying with open science requirements, ensuring that research outputs from Horizon Europe, Euratom, and earlier Framework Programmes are freely accessible, thereby accelerating scientific discovery and innovation.
\nThe EU Open Research Repository serves as a complementary platform to the Open Research Europe (ORE) publishing platform. Open Research Europe focuses on providing a publishing venue for peer-reviewed articles, ensuring that research meets rigorous academic standards. The EU Open Research Repository provides a space for all the other research outputs including data sets, software, posters, and presentations that are out of scope for ORE. This holistic approach enables researchers to not only publish their findings but also share the underlying data and materials that support their work, fostering transparency and reproducibility in the scientific process.
\nCurrently in its pilot phase and set to be fully operational during autumn 2024, the EU Open Research Repository is constantly evolving. Efforts are committed to integrating cutting-edge features, including automated curation checks and FAIR (Findable, Accessible, Interoperable, and Reusable) assistance, to further support the research community. The goal is to provide researchers with a simple goto solution for making their publicly funded research open and as FAIR as possible.
\nThe EU Open Research Repository is funded by the European Union under grant agreement no. 101122956(HORIZON-ZEN). For more information about the project see https://about.zenodo.org/projects/horizon-zen/.
", "title": "EU Open Research Repository (Pilot)", "type": { "id": "organization" }, "website": "https://research-and-innovation.ec.europa.eu" }, "revision_id": 16, "slug": "eu", "theme": { "brand": "horizon", "enabled": true, "style": { "font": { "family": "Arial, sans-serif", "size": "16px", "weight": 600 }, "mainHeaderBackgroundColor": "#FFFFFF", "primaryColor": "#004494", "primaryTextColor": "#FFFFFF", "secondaryColor": "#FFD617", "secondaryTextColor": "#000000", "tertiaryColor": "#e3eefd", "tertiaryTextColor": "#1c5694" } }, "updated": "2024-03-20T06:47:47.577483+00:00" } ], "ids": [ "89ee3a47-3e75-4d1b-b596-2ce9e9ef22ab", "f0a8b890-f97a-4eb2-9eac-8b8a712d3a6c" ] }, "id": "5164244", "pids": {} }, "pids": { "doi": { "identifier": "10.1016/j.neunet.2021.07.021", "provider": "external" }, "oai": { "identifier": "oai:zenodo.org:5164245", "provider": "oai" } }, "revision_id": 3, "stats": { "all_versions": { "data_volume": 120449664.0, "downloads": 76, "unique_downloads": 73, "unique_views": 66, "views": 72 }, "this_version": { "data_volume": 120449664.0, "downloads": 76, "unique_downloads": 73, "unique_views": 66, "views": 72 } }, "status": "published", "updated": "2021-08-10T08:47:28.267518+00:00", "versions": { "index": 1, "is_latest": true } }