{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/lightlda-big-topic-models-on-modest-compute","title":"LightLDA: Big Topic Models on Modest Compute Clusters","arxiv_id":"1412.1576","date":"2014-12-04","proceeding":null,"authors":["Jinhui Yuan","Fei Gao","Qirong Ho","Wei Dai","Jinliang Wei","Xun Zheng","Eric P. Xing","Tie-Yan Liu","Wei-Ying Ma"],"abstract":"When building large-scale machine learning (ML) programs, such as big topic\nmodels or deep neural nets, one usually assumes such tasks can only be\nattempted with industrial-sized clusters with thousands of nodes, which are out\nof reach for most practitioners or academic researchers. We consider this\nchallenge in the context of topic modeling on web-scale corpora, and show that\nwith a modest cluster of as few as 8 machines, we can train a topic model with\n1 million topics and a 1-million-word vocabulary (for a total of 1 trillion\nparameters), on a document collection with 200 billion tokens -- a scale not\nyet reported even with thousands of machines. Our major contributions include:\n1) a new, highly efficient O(1) Metropolis-Hastings sampling algorithm, whose\nrunning cost is (surprisingly) agnostic of model size, and empirically\nconverges nearly an order of magnitude faster than current state-of-the-art\nGibbs samplers; 2) a structure-aware model-parallel scheme, which leverages\ndependencies within the topic model, yielding a sampling strategy that is\nfrugal on machine memory and network communication; 3) a differential\ndata-structure for model storage, which uses separate data structures for high-\nand low-frequency words to allow extremely large models to fit in memory, while\nmaintaining high inference speed; and 4) a bounded asynchronous data-parallel\nscheme, which allows efficient distributed processing of massive data via a\nparameter server. Our distribution strategy is an instance of the\nmodel-and-data-parallel programming model underlying the Petuum framework for\ngeneral distributed ML, and was implemented on top of the Petuum open-source\nsystem. We provide experimental evidence showing how this development puts\nmassive models within reach on a small cluster while still enjoying\nproportional time cost reductions with increasing cluster size, in comparison\nwith alternative options.","url_abs":"http://arxiv.org/abs/1412.1576v1","url_pdf":"http://arxiv.org/pdf/1412.1576v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"lightlda-big-topic-models-on-modest-compute","repo_url":"https://github.com/dmlc/experimental-lda","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"topic-models","task_name":"Topic Models"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}