(Image by Pixabay)
ææã§ããºã£ã¦ãããã®è¨äºã§ããã
ããèªãã ç¬éã«ãããã¼ãããå®å
¨ã«æ´ã®æ±è¥¿ãåãããã¼ã¿åææ¥çã ã¨ã©ãã§ãè¦ããããæ¥çã¤ãã¿ããããã ãªã¼ãã¨ããæã湧ãã¦ãã¦ããã¿ä¸ãã¦ããæ¶ãæ¼ããããã¾ããã§ããï¼åï¼ãã¨ãã軽å£ã¯ãã¦ãããå®éã«ã»ã¼åãå
容ã®æç´ãUSã§ãã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãã¨ãã¦åãã¦ããç¥äººãããç´ã«èãããã¨ãããã®ã§ãå人çã«ãããªã説å¾åã®ãã話ã ãªã¨æãã¾ããã
ãã®è¨äºã®è¥å¹²å«ãªã¨ããã¯ããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ããã¡ã¯ãã¤ã§ã転è·æ´»åããã¦ãããä½æ
ãªãã©ãã®è·å ´ã«è¡ã£ã¦ã以ä¸ã®ã¤ãã¿ãããããã ãã¨ããæ¸ãæ¹ããã¦ããç¹ãããããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãå«ããã¼ã¿åæè·ãå
¨å¡å¸¸ã«è»¢è·æ´»åãã¦ãããã¨ããã¨ãããã«éãã ããã¨æãã¾ãããããããããã¤ãããã§ããã¸åå¼ãã¦ä¸ãããã¨è¨ããããé ·ããããå¾ãªãã®ãã¾ãäºå®ããªã¨ã
ãã¼ã¿åæå°å ¥ã®çæ³å
ã¡ãã£ã¨åã«ãããªè¨äºãæ¸ãã¾ããï¼ãã ãå
ã«ãªã£ãè¨äºã¯ããã«2å¹´ãããé¡ã£ãæã®ãã®ã§ããï¼ããã®ä¸ã§ãç´¹ä»ãã¾ããããå人çã«ã¯ãã¼ã¿åæå°å
¥ã«ã¯7ã¤ã®ã¹ããããããã¨èãã¦ãã¾ãã
- ãã¼ã¿ããã¸ãã¹ã®ãã¨ããã³ã¹ï¼æ ¹æ ï¼ãã¨ããããã¼ã¿ããªãã³ãã«å¤ãã¦ãã
- 次ã«ãã¼ã¿ãéãã¦ãå¯è¦åããã
- ãã¼ã¿ãæ¯è¼ãã¦ãA/Bãã¹ãããã
- A/Bãã¹ãã«ãçµ±è¨å¦çæ¤å®ãã§åç¾æ§ãä¸ãã
- ãå¤å¤éã¢ãã«ããå°å ¥ãã¦åæã«è¤æ°ã®è¦å ã«ã¢ããã¼ããã
- ãã¢ãã«ã«ããäºæ¸¬ï¼æ¨æ¸¬ï¼ãã§æªæ¥ã®ã¢ã¯ã·ã§ã³ãææ決å®ããµãã¼ããã
- ãæ©æ¢°å¦ç¿ãã¨ãã®ãã·ã¹ãã å®è£ ãã«ããèªååãå°å ¥ãã人æã«ãããã¢ãããã¯åæãã¨ãã©ã³ã¹è¯ã使ãåãã
ãããªæãã§é²åãã¦ããã°ãæçµçã«ã¯é常ã«è¯ãæãã§ãã¼ã¿æ´»ç¨ãã§ããããã«ãªãã®ã§ã¯ãªããã¨æãã¾ãã
æ´ã®æ±è¥¿ãåããªãã4ã¤ã®ã¤ãã¿
ããããªããããããªçæ³éãã«ã¯ãããªãã®ããã®ä¸ã®ä¸ãåé ã«æ²ããããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãã¯ã¤ããããè¨äºã§ãè¦åºãã¨ãã¦æãããã¦ãã4ã¤ã®ãã¤ã³ãã以ä¸ã«å¼ç¨ãã¦ããã¾ãã
- Expectation does not match realityï¼ååã»ä¸å¸ããã®æå¾ æãç¾å®ã¨ãããããªãï¼
- Politics reigns supremeï¼ç¤¾å æ¿æ²»ãæåªå ãããï¼
- Youâre the go to person about anything dataï¼ãã¼ã¿ã«é¢ãããã®å ¨ã¦ãæ±ãä½ã§ãå±æ±ããããï¼
- Working in an isolated teamï¼ä»ã®äºæ¥ããå¤ç«ãããã¼ã ã§åããããï¼
å®å ¨ã«è¦ãã¾ãã¾ã§ãããå è¨äºã«ããããã®çç±ã¨äºæ ãæ¸ããã¦ããã®ã§ãããã¤ã¾ãã§ç´¹ä»ãã
ã¾ãã1ã¤ç®ã®ãæå¾ å¤ã³ã³ããã¼ã«ãã«ã¤ãã¦ãããã¯çµæ§è¨ãå¤ããã¦ããåé¡ã ã¨æããã§ãããUSã§ãå ¨ç¶å¤ãããªãã¨ããäºæ ãèªããã¦ãã¾ããè¯éºã«æ©æ¢°å¦ç¿ã"AI"ãé§ä½¿ãã¦åé¡ã解決ãã¦ããã¤ã¡ã¼ã¸ã§ç¾å ´ã«å ¥ã£ã¦ããã¨ãããã«å¾ ã¡åãã¦ããã®ã¯ãã¼ã¿åºç¤ã®æ´åã¨ããã·ã¥ãã¼ãä½ãã«è¿½ãããæ¥ã ãä¸æ¹ã§ãåããããã¡ã¯ããã·ã¥ãã¼ãã«ã¯æºè¶³ãããã®ã®ããã®ç¨åº¦ã®ãã¨ãããããªããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ããéã£ãã¤ããã§ã¯ãªãã£ããã¿ãããªãã¨ãè¨ãåºãããã§ããããã·ã¥ãã¼ãä½ãã ãã§ãã²ã¼ã²ã¼è¨ã£ã¦ãã¨ããã«ããã¡ã®ä¼ç¤¾ã¨ãã¦ã¯AIãé§ä½¿ãã¦å¤¢ã®ãããªã·ã¹ãã ã沢山ä½ãåºãã¦ããããããªåªç§ãªäººæã¨ãã¦åãéã£ãã¯ããªãã ããææãåºã¦ããªãã®ããã¨ããåãããããè¨ãããããã¾ãæ®éã¯ã¤ããã§ããã
2ã¤ç®ã®ã社å æ¿æ²»ããããã¾ãé常ã«å¤ãã®ç¾å ´ããæ¼ãèããããããããã¿ã§ã極端ãªè¨ãæ¹ãããã°ããåãããããæ°ã«å ¥ããããã¨ãä»äºã®ãã¡ããèªåãã¡ã®ãã¼ã¿åæãæ´»ããã¦ãããããããªpipelineãä½ããã¨ãåºæ¥ãªããã°ããã¼ã¿åæã®ä»äºããã¦ãéææã¯å¾ãããªããããåãããããããããã¤ãå½¹ã«ç«ããªããã¨ã°ãããã£ã¦ããã¨è¨ãããç¾½ç®ã«ããªãããã¾ãããã¾ãããã¼ã¿ãåã£ã¦ããã®ã«ã社å æ¿æ²»ãå¿ è¦ãã¨ããã±ã¼ã¹ãããã¾ããä¾ãã°å¿ è¦ãªãã¼ã¿ã®æ ¼ç´å ãèªåãã¡ã®Aæ¬é¨ã¨ã¯å¥ã®Bæ¬é¨ã®DWHã ã£ãå ´åãBæ¬é¨é·ã説å¾ããªãã¨ãã¼ã¿ãè²°ããªãããã§ããããã®Bæ¬é¨é·ãå®ã¯Aæ¬é¨é·ã¨ç¬ç¿ã®ä»²ã§ã©ãã«ããªããªããããã«ç¤¾é·ãéããªãã¨ä½ãã§ããªããã¿ãããªè©±ã¯èãããããªèããªãã£ããããªããã
ããã¦3ã¤ç®ã®ããã¼ã¿ä½ã§ãå±ãããåé¡ãããåãã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãã ããªï¼ãªãSparkãTensorFlowãèªç¶è¨èªå¦çãåããããªï¼ãã¿ãããªãã¨ãè¨ããã¦ãããããã¼ã¿ã¨åã®ã¤ãããããæ¥åãéãããããã¨ã«ãããã¨ãã話ã¯ãããµãéãã¦ãã¦ææã«éããªãã§ããå·¦ã®ã¢ãã¿ã§Stanåãã¦çµ±è¨ã¢ããªã³ã°ãã¦ããä¸æ¹ã§ãå³ã®ã¢ãã¿ã§ã¯Kerasã§æ¨è¦ã·ã¹ãã æ¸ããªãããããªãã¦ã両æ¹ã®ä»äºãä¸æ®µè½ãããä»åº¦ã¯DWHã¸ã®ãã¼ã¿èªã¿è¾¼ã¿ãã³ã±ã¦ããã®ãç´ããªããã°ãããªãæå¥ãå¥ã®äºæ¥é¨ããé ¼ã¾ãã¦ããããã·ã¥ãã¼ããæ´åããªããã°ãããªãã¨ããããå½¹å²åæ ãã§ãã¦ãªãã¨ãã®è¾ºã¯åãªãå°çã«ãªãããã¾ãããåã®æè«ã§ããããã¼ã¿åºç¤æ´åãå°éã«æ å½ãããã¼ã¿ã¢ã¼ããã¯ãè·ãç½®ããã¨ããã®ãããã®è¾ºã®æ¥çäºæ ãè¸ã¾ãã¦ãã¾ãã
æå¾ã«4ã¤ç®ã®ãå¤ç«ãããã¼ã ã§åãããããåé¡ã¯2ã¤ç®ã®ç¤¾å æ¿æ²»ã¨é¢é£ãã¦ãã¾ããè¦ã¯ãã¼ã¿åæã§å¾ãããææããã¡ãã¨ãããã¯ããªãæ¬çªã·ã¹ãã ãªãã«åæ ãããããªããã°æå³ããªãã®ã«ãç¾å®ã«ã¯ãã¼ã¿åæå°éã®ãã¼ã ãåå¥ã«çµç¹ããã¦ãã¦åæä»äºã¯ããã ãã§è¡ãããã®ã«å¯¾ãã¦ããããã®ææãåæ ããããã¨ããã¨å®éã®ãããã¯ããæ¬çªã·ã¹ãã ã«é¢ããç¥èãã¹ãã«ãå¿ è¦ã§ããã¾ã§å°éã§ããªãï¼ãããã¯ãããããªãã°ããã«ãããã¯ããæ¬çªã·ã¹ãã ã®ãã¼ã ã«ç¸æã«ãããªãããããªãã°ããããªãç«æ´¾ãªã¢ã«ã´ãªãºã ãçµã¿è¾¼ã¿ãããªãèªåã§ã³ã¼ãæ¸ãã¦ã³ã¼ãã¬ãã¥ã¼éãã¦ã³ããããããã¨ãè¦æ±ãããï¼ã¨ãã話ã§ããããã¾ãè¯ãèã話ã§ããå®éã®äºæ¥é¨éã¨ã®é£æºãå¼±ãã¨ãã©ããã¦ããã¼ã¿åæã®ç価ãåæ ãããããªãã¨ããã®ã¯ã¾ã¾ããã¾ãã
ãã¡ãããä¸ã®ä¸ã«ã¯ãããã®ã¤ãã¿ãå
¨ã¦ã¯ãªã¢ãã夢ã®ãããªç¾å ´ããªãããã§ã¯ããã¾ããã*1ãæ¯è¼ã®åé¡ã§è¨ãã°ãã¯ãããããã¨ããã¯ç¨å°ã§ãå¤ãã®ç¾å ´ã¯ãããã®ã¤ãã¿ã®ã©ãã1ã¤ä»¥ä¸ï¼ãã£ããããã¨4ã¤å
¨ã¦ï¼ãæ±ãã¦ããã¨è¨ã£ã¦è¯ããã¨æãã¾ãããããªãã¨ãå°ãã§ãã¤ãã¿ã®å°ãªãç¾å ´ãæ±ãã¦ãã¼ã¿ãµã¤ã¨ã³ãã£ã¹ããã¡ã¯æµæµªã®æ°ã«ãªããã¡ã ãã¨ããã®ããã®è¨äºã®è¨ãããã£ããã¨ã®ããã§ãã
ãã¼ã¿ãµã¤ã¨ã³ã¹ã®ä¸é½åãªçå®
以ä¸ã§ç´¹ä»ããã®ã¯ã©ã¡ããã¨ããã¨ãä»äºã¨ãã¦ã®ããã¼ã¿åæè·ã®ã¤ãã¿ã¨ããã話ã§ãããã¡ãªã¿ã«ã¹ãã«é¢ã§ã®ã¤ãã¿ã«ã¤ãã¦ã¯"The Inconvenient Truth About Data Science"ã¨ããå°ãã¿ãæ¢ã«ããã¾ãã¦ã
- Data is never clean.ï¼ãã¼ã¿ã¯ã©ãã綺éºã§ã¯ãªãï¼
- You will spend most of your time cleaning and preparing data.ï¼å¤§åã®æéã¯åå¦çã«è²»ããããï¼
- 95% of tasks do not require deep learning.ï¼95%ã®ä»äºã¯Deep Learningãå¿ è¦ã¨ããªãï¼
- In 90% of cases generalized linear regression will do the trick.
ï¼GLMã®90%ã¯åãªãããªãã¯ï¼ï¼90%ã®ã±ã¼ã¹ã§GLMã¯ãã¾ãããï¼*2- Big Data is just a tool.ï¼ããã°ãã¼ã¿ã¯ãã ã®éå ·ï¼
- You should embrace the Bayesian approach.ï¼ãã¤ã¸ã¢ã³ã«å¸°ä¾ããï¼
- No one cares how you did it.ï¼ã©ãããããæ¹ããããã誰ãæ°ã«ãããªãï¼*3
- Academia and business are two different worlds.ï¼å¦è¡çã¨ç£æ¥çã¨ã¯2ã¤ã®ç°ãªãä¸çã ï¼
- Presentation is key - be a master of Power Point.ï¼ãã¬ã¼ã³ã¯éè¦ã ï¼PowerPointãã¹ã¿ã¼ã«ãªããï¼
- All models are false, but some are useful.ï¼å ¨ã¦ã®ã¢ãã«ã¯åã ãã ãä¸ã«ã¯å½¹ç«ã¤ãã®ãããï¼*4
- There is no fully automated Data Science. You need to get your hands dirty.ï¼å ¨èªååããããã¼ã¿ãµã¤ã¨ã³ã¹ãªãã¦ãã®ã¯ãªããèªãæãæ±ãã¦åãï¼
ã³ã³ãã¨ãã¦ã¯å¤§ä½åãã ãªã¨å人çã«ã¯æãã¾ãããããããæ´ã®æ±è¥¿ãåããªãæ§ã ãªç¨®é¡ã®ã¤ãã¿ãæ±ããªããããã¼ã¿åæè·ã®äººã ã¯æ¥ã ã®æ¥åã¨æ ¼éããªããã°ãããªãããã§ãããããªã¤ãã¿ã ããã ããããæµæµªã®æ°ã«ãªãã¨ããã®ãã¾ãéçããªã¨ã
ã¨ãããããªããã©ãå®æ ãç解ããä¸ã§ãæè¿å種ã¡ãã£ã¢è¨äºã§å§ä¼ããã¦ãããã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãè²æã³ã³ã½ã¼ã·ã¢ã ï¼ã®ãããªãã®ãæ§æ³ããã¦ããã®ã§ããã°åã°ãããã¨ã ã¨æãã®ã§ãããå®éã®ã¨ããã¯ã©ããªã®ããªãã¨ããå±æ§ãè¦ãã¾ããããç¾å ´ããã¯ä»¥ä¸ã§ãã
*1:åã ã¨æãããããããã¾ããããä¼èã®ç¯å²ã§ã¯ãããã夢ã®ãããªç¾å ´ã®åå¨èªä½ã¯ç¢ºèªããã¦ãã¾ã
*2:ãã¡ã http://b.hatena.ne.jp/entry/361518076/comment/mbr ã®éãã§ãorz id:mbrããããæææé£ããããã¾ãã
*3:Excelã§çä»ããããé«åº¦ãªã¢ãã«ãé§ä½¿ãã¦çä»ãããããæ®éã®äººã¯åºå¥ãã¤ããªãã¨ããã話