NetPrune: A sparklines visualization for network pruning
Current deep learning approaches are cutting-edge methods for solving classification tasks. Arising transfer learning techniques allows applying large generic model to simple tasks whereas simpler models could be used. Large models raise the major problem of their memory consumption and processor us...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-06-01
|
| Series: | Visual Informatics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2468502X23000141 |
| _version_ | 1797796386739060736 |
|---|---|
| author | Luc-Etienne Pommé Romain Bourqui Romain Giot Jason Vallet David Auber |
| author_facet | Luc-Etienne Pommé Romain Bourqui Romain Giot Jason Vallet David Auber |
| author_sort | Luc-Etienne Pommé |
| collection | DOAJ |
| description | Current deep learning approaches are cutting-edge methods for solving classification tasks. Arising transfer learning techniques allows applying large generic model to simple tasks whereas simpler models could be used. Large models raise the major problem of their memory consumption and processor usage and lead to a prohibitive ecological footprint. In that paper, we present a novel visual analytics approach to interactively prune those networks and thus limit that issue. Our technique leverages a novel sparkline matrix visualization technique as well as a novel local metric which evaluates the discriminatory power of a filter to guide the pruning process and make it interpretable. We assess the well- founded of our approach through two realistic case studies and a user study. For both of them, the interactive refinement of the model led to a significantly smaller model having similar prediction accuracy than the original one. |
| first_indexed | 2024-03-13T03:32:19Z |
| format | Article |
| id | doaj.art-531f15f6745e4cb48a08ccb8624d5838 |
| institution | Directory Open Access Journal |
| issn | 2468-502X |
| language | English |
| last_indexed | 2024-03-13T03:32:19Z |
| publishDate | 2023-06-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Visual Informatics |
| spelling | doaj.art-531f15f6745e4cb48a08ccb8624d58382023-06-24T05:18:22ZengElsevierVisual Informatics2468-502X2023-06-01728599NetPrune: A sparklines visualization for network pruningLuc-Etienne Pommé0Romain Bourqui1Romain Giot2Jason Vallet3David Auber4Corresponding author.; Univ. Bordeaux, CNRS, Bordeaux INP, INRIA, LaBRI, UMR 5800, F-33400 Talence, FranceUniv. Bordeaux, CNRS, Bordeaux INP, INRIA, LaBRI, UMR 5800, F-33400 Talence, FranceUniv. Bordeaux, CNRS, Bordeaux INP, INRIA, LaBRI, UMR 5800, F-33400 Talence, FranceUniv. Bordeaux, CNRS, Bordeaux INP, INRIA, LaBRI, UMR 5800, F-33400 Talence, FranceUniv. Bordeaux, CNRS, Bordeaux INP, INRIA, LaBRI, UMR 5800, F-33400 Talence, FranceCurrent deep learning approaches are cutting-edge methods for solving classification tasks. Arising transfer learning techniques allows applying large generic model to simple tasks whereas simpler models could be used. Large models raise the major problem of their memory consumption and processor usage and lead to a prohibitive ecological footprint. In that paper, we present a novel visual analytics approach to interactively prune those networks and thus limit that issue. Our technique leverages a novel sparkline matrix visualization technique as well as a novel local metric which evaluates the discriminatory power of a filter to guide the pruning process and make it interpretable. We assess the well- founded of our approach through two realistic case studies and a user study. For both of them, the interactive refinement of the model led to a significantly smaller model having similar prediction accuracy than the original one.http://www.sciencedirect.com/science/article/pii/S2468502X23000141Explainable pruningGuided Fine-tuningVisualizationDeep learning |
| spellingShingle | Luc-Etienne Pommé Romain Bourqui Romain Giot Jason Vallet David Auber NetPrune: A sparklines visualization for network pruning Visual Informatics Explainable pruning Guided Fine-tuning Visualization Deep learning |
| title | NetPrune: A sparklines visualization for network pruning |
| title_full | NetPrune: A sparklines visualization for network pruning |
| title_fullStr | NetPrune: A sparklines visualization for network pruning |
| title_full_unstemmed | NetPrune: A sparklines visualization for network pruning |
| title_short | NetPrune: A sparklines visualization for network pruning |
| title_sort | netprune a sparklines visualization for network pruning |
| topic | Explainable pruning Guided Fine-tuning Visualization Deep learning |
| url | http://www.sciencedirect.com/science/article/pii/S2468502X23000141 |
| work_keys_str_mv | AT lucetiennepomme netpruneasparklinesvisualizationfornetworkpruning AT romainbourqui netpruneasparklinesvisualizationfornetworkpruning AT romaingiot netpruneasparklinesvisualizationfornetworkpruning AT jasonvallet netpruneasparklinesvisualizationfornetworkpruning AT davidauber netpruneasparklinesvisualizationfornetworkpruning |