{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2024,8,7]],"date-time":"2024-08-07T07:34:32Z","timestamp":1723016072345},"publisher-location":"California","reference-count":0,"publisher":"International Joint Conferences on Artificial Intelligence Organization","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2020,7]]},"abstract":"<jats:p>In this paper we present a novel mechanism to\n\nget explanations that allow to better understand\n\nnetwork predictions when dealing with sequential\n\ndata. Specifically, we adopt memory-based networks\n\n\u2014 Differential Neural Computers \u2014 to exploit \n\ntheir capability of storing data in memory and\n\nreusing it for inference. By tracking both the memory \n\naccess at prediction time, and the information\n\nstored by the network at each step of the input\n\nsequence, we can retrieve the most relevant input\n\nsteps associated to each prediction. We validate\n\nour approach (1) on a modified T-maze, which is a\n\nnon-Markovian discrete control task evaluating an\n\nalgorithm\u2019s ability to correlate events far apart in\n\nhistory, and (2) on the Story Cloze Test, which is\n\na commonsense reasoning framework for evaluating \n\nstory understanding that requires a system to\n\nchoose the correct ending to a four-sentence story.\n\nOur results show that we are able to explain agent\u2019s\n\ndecisions in (1) and to reconstruct the most relevant\n\nsentences used by the network to select the story\n\nending in (2). Additionally, we show not only that\n\nby removing those sentences the network prediction\n\nchanges, but also that the same are sufficient to\n\nreproduce the inference.<\/jats:p>","DOI":"10.24963\/ijcai.2020\/278","type":"proceedings-article","created":{"date-parts":[[2020,7,8]],"date-time":"2020-07-08T12:12:10Z","timestamp":1594210330000},"page":"2006-2013","source":"Crossref","is-referenced-by-count":4,"title":["Explainable Inference on Sequential Data via Memory-Tracking"],"prefix":"10.24963","author":[{"given":"Biagio","family":"La Rosa","sequence":"first","affiliation":[{"name":"Sapienza University of Rome, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Roberto","family":"Capobianco","sequence":"additional","affiliation":[{"name":"Sapienza University of Rome, Italy"},{"name":"Sony AI"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Daniele","family":"Nardi","sequence":"additional","affiliation":[{"name":"Sapienza University of Rome, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"10584","event":{"number":"28","sponsor":["International Joint Conferences on Artificial Intelligence Organization (IJCAI)"],"acronym":"IJCAI-PRICAI-2020","name":"Twenty-Ninth International Joint Conference on Artificial Intelligence and Seventeenth Pacific Rim International Conference on Artificial Intelligence {IJCAI-PRICAI-20}","start":{"date-parts":[[2020,7,11]]},"theme":"Artificial Intelligence","location":"Yokohama, Japan","end":{"date-parts":[[2020,7,17]]}},"container-title":["Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence"],"original-title":[],"deposited":{"date-parts":[[2020,7,9]],"date-time":"2020-07-09T02:14:15Z","timestamp":1594260855000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ijcai.org\/proceedings\/2020\/278"}},"subtitle":[],"proceedings-subject":"Artificial Intelligence Research Articles","short-title":[],"issued":{"date-parts":[[2020,7]]},"references-count":0,"URL":"https:\/\/doi.org\/10.24963\/ijcai.2020\/278","relation":{},"subject":[],"published":{"date-parts":[[2020,7]]}}}