å®è·µXAI[説æå¯è½ãªAIï¼½ æ©æ¢°å¦ç¿ã®äºæ¸¬ã説æããããã®Pythonã³ã¼ãã£ã³ã°
3,960å
1,980å
(1,800å+ç¨)
ãç¹ä¾¡ã
ã注æãæ¬æ¸ã®EPUBçã¯åºå®ã¬ã¤ã¢ã¦ãåã«ãªã£ã¦ããã¾ããæåã®å¤§ããã®å¤æ´ãæ¤ç´¢ãå¼ç¨ãªã©ã¯ã使ãããã ãã¾ãããç»é¢ã®å¤§ãã端æ«ã§ãå©ç¨ãã ããã
é¢é£ãµã¤ã
æ¬æ¸ã®é¢é£ãã¼ã¸ãç¨æããã¦ãã¾ãã
å 容紹ä»
XAI Pythonã©ã¤ãã©ãªã§äºæ¸¬çµæã説æããã©ãã¯ããã¯ã¹ããã°ã©ã¹ããã¯ã¹ã¸ã
ãã¸ãã¹ä¸ã®ææ決å®ã«ã¤ãªããæ©æ¢°å¦ç¿ã®äºæ¸¬ã«ã¯ã解éã»èª¬æãæ±ãããã¾ããããã§ããããã解éã»èª¬æãè¡ãããã®ææ³ãç´¹ä»ãã¦ããã¾ãã
ã¾ããã¢ãã«ã®èª¬æå¯è½æ§ã¨è§£éå¯è½æ§ã®å ¨ä½åãå«ççã«èæ ®ãã¹ãç¹ãçæãããäºæ¸¬ã®ãã¤ã¢ã¹ã«ã¤ãã¦èª¬æãã¾ãã
次ã«ãLIMEãSHAPãSkaterãELI5ãskope-rulesã¨ãã£ãPythonã©ã¤ãã©ãªã使ã£ã¦ãã¢ãã«ããªããã®ããã«äºæ¸¬ããã®ããæ¢ã£ã¦ããã¾ãã
äºæ¸¬ã¢ãã«ã¨ãã¦ãç·å½¢ã»éç·å½¢ã¢ãã«ã®ã»ããã¢ã³ãµã³ãã«ã¢ãã«ãæç³»åã¢ãã«ãèªç¶è¨èªå¦çããã£ã¼ãã©ã¼ãã³ã°ãã³ã³ãã¥ã¼ã¿ã¼ãã¸ã§ã³ãåãä¸ãã¾ãã
æ¬æ¸ã¯è§£éã»èª¬æã®ããã®æ¹æ³ãå æ¬çã«åãä¸ãã¦ãããæ©æ¢°å¦ç¿ãå®éã®ç¾å ´ã§æ´»ç¨ããæ¹ã«ãã²æã«åã£ã¦ããã ãããä¸åã§ãã
âPythonã®ææ³ã»ã©ã¤ãã©ãªãæ©æ¢°å¦ç¿ã«ã¤ãã¦æ¨æºçãªç¥èãç解ãã¦ããå¿ è¦ãããã¾ãã
æ¸èªæ å ±
- èè : Pradeepta Mishra(è), æ ªå¼ä¼ç¤¾ã¯ã¤ã¼ã(訳)
- çºè¡æ¥: 2023-06-20 (ç´æ¸ç±ççºè¡æ¥: 2023-06-20)
- æçµæ´æ°æ¥: 2023-06-20
- ãã¼ã¸ã§ã³: 1.0.0
- ãã¼ã¸æ°: 322ãã¼ã¸(PDFçæç®)
- 対å¿ãã©ã¼ããã: PDF, EPUB
- åºç社: ã¤ã³ãã¬ã¹
対象èªè
ãã¼ã¿ãµã¤ã¨ã³ãã£ã¹ããAIããã¸ã§ã¯ãã®é¢ä¿è ãéçºè
èè ã«ã¤ãã¦
Pradeepta Mishra
ã¤ã³ããæ ç¹ã¨ããå¤å½ç±ä¼æ¥ã®ITãµã¼ãã¹å ¼ã³ã³ãµã«ãã£ã³ã°ä¼ç¤¾Lï¼T Infotechã®AIãã¼ã¿ãããã¯ãé¨éä¸å¸ããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ããè¨ç®è¨èªå¦ã¨ãã¹ãã¼ããæ©æ¢°å¦ç¿ã»æ·±å±¤å¦ç¿ã®ã¨ãã¹ãã¼ããããªã大è¦æ¨¡ãªã°ã«ã¼ããçããã以åã«ã¯ãAnalytics India Magazineã®ãIndia's Top - 40 Under 40 DataScientistsãã«é¸åºããããã¾ãããã¼ã¿ãµã¤ã¨ã³ã¹ãAIã«é¢ãã500以ä¸ã®æè¡è¬æ¼ãããã¾ãã¾ãªå¤§å¦ãæè¡æ©é¢ãã³ãã¥ããã£ãªã©ã§è¡ãã
æ ªå¼ä¼ç¤¾ã¯ã¤ã¼ã
ç®æ¬¡
ã¾ããã
ç®æ¬¡
第1ç« ã¢ãã«ã®èª¬æå¯è½æ§ã¨è§£éå¯è½æ§
- 1.1 ãã¬ã¼ã ã¯ã¼ã¯ã®ç¢ºç«
- 1.2 AI
- 1.2.1 XAIã®å¿ è¦æ§
- 1.2.2 説æå¯è½æ§ã¨è§£éå¯è½æ§
- 1.2.3 説æï¼è§£éï¼å¯è½æ§ã®ç¨®é¡
- 1.2.4 ã¢ãã«ã®èª¬æå¯è½æ§ã®ããã®ãã¼ã«
- 1.2.5 SHAP
- 1.2.6 LIME
- 1.2.7 ELI5
- 1.2.8 Skater
- 1.2.9 skope-rules
- 1.2.10 æ©æ¢°å¦ç¿ã®ããã®XAIã®ææ³
- 1.2.11 XAIäºæã®ã¢ãã«
- 1.2.12 XAIã¨è²¬ä»»ããAI
- 1.2.13 XAIã®è©ä¾¡
- 1.3 ã¾ã¨ã
第2ç« AIã®å«çãåè¦ãä¿¡é ¼æ§
- 2.1 éç¿ï¼AIã®å«ç
- 2.2 AIã®åè¦
- 2.3 ãã¼ã¿ã®ãã¤ã¢ã¹
- 2.4 ã¢ã«ã´ãªãºã ã®ãã¤ã¢ã¹
- 2.5 ãã¤ã¢ã¹ãæ¸ããããã»ã¹
- 2.6 解éã®ãã¤ã¢ã¹
- 2.7 è¨ç·´ã®ãã¤ã¢ã¹
- 2.8 AIã®ä¿¡é ¼æ§
- 2.9 ã¾ã¨ã
第3ç« ç·å½¢ã¢ãã«ã®èª¬æå¯è½æ§
- 3.1 ç·å½¢ã¢ãã«
- 3.2 ç·å½¢å帰
- 3.3 çºçããå¯è½æ§ãããåé¡ã¨VIF
- 3.3.1 æçµçãªã¢ãã«
- 3.3.2 ã¢ãã«ã®èª¬æå¯è½æ§
- 3.4 æ©æ¢°å¦ç¿ã¢ãã«ã¸ã®ä¿¡é ¼ï¼SHAP
- 3.4.1 æ©æ¢°å¦ç¿ã¢ãã«ã§ã®å±æçãªèª¬æã¨åã ã®äºæ¸¬å¤
- 3.4.2 æ©æ¢°å¦ç¿ã¢ãã«ã§ã®å¤§åçãªèª¬æã¨å ¨ä½çãªäºæ¸¬å¤
- 3.5 LIMEã«ãã説æã¨æ©æ¢°å¦ç¿ã¢ãã«
- 3.6 Skaterã«ãã説æã¨æ©æ¢°å¦ç¿ã¢ãã«
- 3.7 ELI5ã«ãã説æã¨æ©æ¢°å¦ç¿ã¢ãã«
- 3.8 ãã¸ã¹ãã£ãã¯å帰
- 3.8.1 解é
- 3.8.2 LIMEã®æ¨è«
- 3.9 ã¾ã¨ã
第4ç« éç·å½¢ã¢ãã«ã®èª¬æå¯è½æ§
- 4.1 éç·å½¢ã¢ãã«
- 4.2 決å®æ¨ã®èª¬æ
- 4.3 決å®æ¨ã¢ãã«ã®ãã¼ã¿ãæºåãã
- 4.3.1 決å®æ¨ã¢ãã«ãä½æãã
- 4.4 決å®æ¨ï¼SHAP
- 4.5 SHAPã使ã£ãPDP
- 4.6 scikit-learnã使ã£ãPDP
- 4.7 éç·å½¢ã¢ãã«ã®èª¬æï¼LIME
- 4.8 éç·å½¢ã¢ãã«ã®èª¬æï¼skope-rules
- 4.9 ã¾ã¨ã
第5ç« ã¢ã³ãµã³ãã«ã¢ãã«ã®èª¬æå¯è½æ§
- 5.1 ã¢ã³ãµã³ãã«ã¢ãã«
- 5.1.1 ã¢ã³ãµã³ãã«ã¢ãã«ã®ç¨®é¡
- 5.2 ã¢ã³ãµã³ãã«ã¢ãã«ã使ãã®ã¯ãªãã
- 5.3 ã¢ã³ãµã³ãã«ã¢ãã«ã§SHAPã使ã
- 5.4 InterpretMLã使ã£ã¦ãã¼ã¹ãã£ã³ã°ã¢ãã«ã説æãã
- 5.5 ã¢ã³ãµã³ãã«åé¡ã¢ãã«ï¼SHAP
- 5.6 SHAPã使ã£ã¦CatBoostã¢ãã«ã説æãã
- 5.7 SHAPã使ã£ã¦CatBoostãã¼ã¹ã®å¤ã¯ã©ã¹åé¡ã¢ãã«ã説æãã
- 5.8 SHAPã使ã£ã¦LightGBMã¢ãã«ã説æãã
- 5.9 ã¾ã¨ã
第6ç« æç³»åã¢ãã«ã®èª¬æå¯è½æ§
- 6.1 æç³»åã¢ãã«
- 6.2 ã©ã®ã¢ãã«ãé©åããç¥ã
- 6.2.1 äºæ¸¬ã®ããã®æ¦ç¥
- 6.2.2 äºæ¸¬ã®ä¿¡é ¼åºé
- 6.3 ä¿¡é ¼ã¯ã©ããªãï¼
- 6.4 æç³»åã¢ãã«ï¼LIME
- 6.5 ã¾ã¨ã
第7ç« èªç¶è¨èªå¦çã®èª¬æå¯è½æ§
- 7.1 èªç¶è¨èªå¦çã®ã¿ã¹ã¯
- 7.2 ããã¹ãåé¡ã®èª¬æå¯è½æ§
- 7.3 ããã¹ãåé¡ç¨ã®ãã¼ã¿ã»ãã
- 7.4 ELI5ã使ã£ã¦èª¬æãã
- 7.5 å±æçãªèª¬æã«å¯¾ããç¹å¾´éã®éã¿ãè¨ç®ãã
- 7.5.1 å±æçãªèª¬æï¼ä¾1
- 7.5.2 å±æçãªèª¬æï¼ä¾2
- 7.5.3 ã¹ãããã¯ã¼ããåãé¤ããå¾ã®èª¬æ
- 7.6 n-gramãã¼ã¹ã®ããã¹ãåé¡
- 7.7 è¤æ°ã©ãã«ã®å¤ã¯ã©ã¹ããã¹ãåé¡ã®èª¬æå¯è½æ§
- 7.7.1 å±æçãªèª¬æï¼ä¾1
- 7.7.2 å±æçãªèª¬æï¼ä¾2
- 7.7.3 å±æçãªèª¬æï¼ä¾3
- 7.8 ã¾ã¨ã
第8ç« What-Ifã·ããªãªã使ã£ãã¢ãã«ã®å ¬å¹³æ§
- 8.1 WITã¨ã¯ä½ã
- 8.2 WITã®ã¤ã³ã¹ãã¼ã«ã¨æ´»ç¨
- 8.2.1 è©ä¾¡ææ¨
- 8.3 ã¾ã¨ã
第9ç« ãã£ã¼ãã©ã¼ãã³ã°ã¢ãã«ã®èª¬æå¯è½æ§
- 9.1 ãã£ã¼ãã©ã¼ãã³ã°ã¢ãã«ã説æãã
- 9.2 ãã£ã¼ãã©ã¼ãã³ã°ã§SHAPã使ã
- 9.2.1 Deep SHAPã使ã
- 9.2.2 ç»ååé¡ã§SHAP DeepExplainerã使ãï¼ä¾1
- 9.2.3 ç»ååé¡ã§SHAP DeepExplainerã使ãï¼ä¾2
- 9.2.4 表形å¼ãã¼ã¿ã§SHAP DeepExplainerã使ã
- 9.3 ã¾ã¨ã
第10ç« XAIã¢ãã«ã®åå®ä»®æ³èª¬æ
- 10.1 åå®ä»®æ³èª¬æã¨ã¯ä½ã
- 10.2 åå®ä»®æ³èª¬æãå®è£ ãã
- 10.3 Alibiã使ã£ãåå®ä»®æ³èª¬æ
- 10.4 å帰ã¿ã¹ã¯ã§ã®åå®ä»®æ³èª¬æ
- 10.5 ã¾ã¨ã
第11ç« æ©æ¢°å¦ç¿ã§ã®å¯¾æ¯ç説æ
- 11.1 æ©æ¢°å¦ç¿ã§ã®å¯¾æ¯ç説æã¨ã¯ä½ã
- 11.2 Alibiã使ã£ãã¢ãã«ã®å¯¾æ¯ç説æ
- 11.2.1 å ã®ç»åã¨ãªã¼ãã¨ã³ã³ã¼ããçæããç»åãæ¯è¼ãã
- 11.2.2 表形å¼ãã¼ã¿ã§ã®CEM
- 11.3 ã¾ã¨ã
第12ç« äºæ¸¬ä¸å¤æ§ã®ç¹å®ã«ããã¢ãã«ä¸å¯ç¥ã®èª¬æ
- 12.1 ã¢ãã«ä¸å¯ç¥ã¨ã¯ä½ã
- 12.2 ã¢ã³ã«ã¼ã¨ã¯ä½ã
- 12.3 Alibiã使ã£ãã¢ã³ã«ã¼èª¬æ
- 12.4 ããã¹ãåé¡ã§AnchorTextã使ã
- 12.5 ç»ååé¡ã§AnchorImageã使ã
- 12.6 ã¾ã¨ã
第13ç« ã«ã¼ã«ãã¼ã¹ã®ã¨ãã¹ãã¼ãã·ã¹ãã ã§ã®ã¢ãã«ã®èª¬æå¯è½æ§
- 13.1 ã¨ãã¹ãã¼ãã·ã¹ãã ã¨ã¯ä½ã
- 13.1.1 ååãé£éã¨å¾ãåãé£é
- 13.2 scikit-learnã使ã£ãã«ã¼ã«æ½åº
- 13.3 ã«ã¼ã«ãã¼ã¹ã®ã·ã¹ãã ãå¿ è¦ãªçç±
- 13.4 ã¨ãã¹ãã¼ãã·ã¹ãã ã®èª²é¡
- 13.5 ã¾ã¨ã
第14ç« ã³ã³ãã¥ã¼ã¿ãã¸ã§ã³ã§ã®ã¢ãã«ã®èª¬æå¯è½æ§
- 14.1 ç»åãã¼ã¿ã§ã®èª¬æå¯è½æ§
- 14.1.1 ã³ã³ãã¥ã¼ã¿ãã¸ã§ã³ã§ã¢ã³ã«ã¼ç»åã使ã
- 14.2 å¾é ç©åæ³
- 14.3 ã¾ã¨ã