[機械å¦ç¿’] LDAã®ã‚³ãƒ¼ãƒ‰ã‚’書ã„ã¦ã¿ãŸ - tsubosakaã®æ—¥è¨˜æ˜”書ã„ãŸã“ã¨ãŒã‚ã£ãŸã‘ã©ã€ã©ã“ã‹ã«ã„ã£ã¦ã—ã¾ã£ãŸã®ã§ã‚‚ã†ä¸€åº¦æ›¸ã„ã¦ã¿ãŸã€‚推論方法ã«ã¯ã‚®ãƒ–スサンプリングã¨å¤‰åˆ†ãƒ™ã‚¤ã‚ºã®2ã¤ãŒã‚ã‚‹ã‘ã©ã€å°Žå‡ºã‚‚実装もより楽ãªcollapsed gibbs sampling(Griffiths and Steyvers, PNAS, 2004)ã®æ–¹ã‚’採用。 Token.java package lda; public class Token { public int docId; public int wordId; public Token(int d , int w){ docId = d; wordId = w; } } LDA.java package lda; import java.util.*; public class LDA { int D; // number of document int K; // number of topic int
![[機械å¦ç¿’] LDAã®ã‚³ãƒ¼ãƒ‰ã‚’書ã„ã¦ã¿ãŸ - tsubosakaã®æ—¥è¨˜](https://cdn-ak-scissors.b.st-hatena.com/image/square/496431e556868bdf2bfbac690d83310a1e4ba258/height=288;version=1;width=512/https%3A%2F%2Fogimage.blog.st-hatena.com%2F17680117127124028144%2F17680117127124031279%2F1557332459)
{{#tags}}- {{label}}
{{/tags}}