{"url":"/method/compressed-memory","slug":"compressed-memory","name":"Compressed Memory","full_name":"Compressed Memory","full_name_withheld":false,"description_markdown":"**Compressed Memory** is a secondary FIFO memory component proposed as part of the [Compressive Transformer](https://paperswithcode.com/method/compressive-transformer) model. The Compressive [Transformer](https://paperswithcode.com/method/transformer) keeps a fine-grained memory of past activations, which are then compressed into coarser compressed memories. \r\n\r\nFor choices of compression functions $f\\_{c}$ the authors consider (1) max/mean pooling, where the kernel and stride is set to the compression rate $c$; (2) 1D [convolution](https://paperswithcode.com/method/convolution) also with kernel & stride set to $c$; (3) dilated convolutions; (4) *most-used* where the memories are sorted by their average attention (usage) and the most-used are preserved.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Compressive Transformers for Long-Range Sequence Modelling","paper":"/paper/compressive-transformers-for-long-range-1","first_author":"Jack W. Rae","n_authors":4,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/compressive-transformers-for-long-range-1"},"source":{"url":"https://arxiv.org/abs/1911.05507v1","title":"Compressive Transformers for Long-Range Sequence Modelling","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/lucidrains/compressive-transformer-pytorch/blob/a73d146ee7bc9e47cc6c525f645a5bbb583537ad/compressive_transformer_pytorch/compressive_transformer_pytorch.py#L252","code_snippet_url_on_a_code_host":true,"categories":[{"area":"General","area_id":"general","collection":"Miscellaneous Components","url":"/methods/category/miscellaneous-components","pwc_aliases":[]}],"n_papers_tagged":5,"archive_num_papers":5,"papers_newest_first":[{"paper":null,"title":"EdgeInfinite: A Memory-Efficient Infinite-Context Transformer for Edge Devices","date":"2025-03-28","arxiv_id":"2503.22196","n_code_links":0,"syntology":null},{"paper":null,"title":"LoMA: Lossless Compressed Memory Attention","date":"2024-01-16","arxiv_id":"2401.09486","n_code_links":0,"syntology":null},{"paper":null,"title":"Joint Modeling of Feature, Correspondence, and a Compressed Memory for Video Object Segmentation","date":"2023-08-25","arxiv_id":"2308.13505","n_code_links":0,"syntology":null},{"paper":null,"title":"DCT: Dynamic Compressive Transformer for Modeling Unbounded Sequence","date":"2021-10-10","arxiv_id":"2110.04821","n_code_links":0,"syntology":null},{"paper":"/paper/compressive-transformers-for-long-range-1","title":"Compressive Transformers for Long-Range Sequence Modelling","date":"2019-11-13","arxiv_id":"1911.05507","n_code_links":6,"syntology":{"ran":3,"of":11,"unverified":8,"pointer_only":0}}],"papers_shown":5,"tasks":[{"task":null,"name":"GPU","papers":1},{"task":"/task/language-modelling","name":"Language Modelling","papers":1},{"task":"/task/semantic-segmentation","name":"Semantic Segmentation","papers":1},{"task":"/task/sentence","name":"Sentence","papers":1},{"task":"/task/video-object-segmentation","name":"Video Object Segmentation","papers":1},{"task":"/task/video-semantic-segmentation","name":"Video Semantic Segmentation","papers":1}],"tasks_shown":6,"n_tasks":6,"usage_by_year":[{"year":"2019","papers":1},{"year":"2021","papers":1},{"year":"2023","papers":1},{"year":"2024","papers":1},{"year":"2025","papers":1}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/compressed-memory"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}