浠婃棩Nature: 浜哄伐鏅鸿兘浠?鍒?, 鏃犲笀鑷€氬畬鐖嗛樋娉曠嫍100-0|娣卞害瑙ｆ瀽
杩欓噷鏄爣棰樹竴h1鍗犱綅鏂囧瓧
鍙戝竷鏃堕棿锛
2017-10-21
鍏抽敭璇嶏細
銆€銆€鍘诲勾锛屾湁涓皬瀛╄閬嶄汉涓栨墍鏈夌殑妫嬭氨锛岃緵鍕ゆ墦璋憋紝鑻︽€濆啣鎯筹紝妫嬭壓绮捐繘锛?-1鎵撹触涓栫晫鍐犲啗鏉庝笘鐭筹紝浠庢浜洪棿鏃犳晫鎵嬨€備粬鐨勫悕瀛楀彨闃挎硶鐙椼€侟/div>
銆€銆€浠婂勾锛屼粬鐨勫紵寮熷彧闈犱竴鍓鐩樺拰榛戠櫧涓ゅ瓙锛屾病鐪嬭繃涓€涓璋憋紝涔熸病鏈変竴涓汉鎸囩偣锛屼粠闆跺紑濮嬶紝鑷ū鑷箰锛岃嚜宸卞弬鎮燂紝100-0鎵撹触鍝ュ摜闃挎硶鐙椼€備粬鐨勫悕瀛楀彨闃挎硶鍏冦€侟/div>
銆€銆€DeepMind杩欓」浼熷ぇ鐨勭獊鐮达紝浠婂ぉ浠astering the game of Go without human knowledge涓洪锛屽彂琛ㄤ簬Nature锛屽紩璧疯桨鍔ㄣ€傜煡绀剧壒閭€鍥藉唴澶栧嚑浣嶄汉宸ユ櫤鑳戒笓瀹讹紝缁欎簣娣卞害瑙ｆ瀽鍜岀偣璇勩€傛枃鏈湁DeepMind David Silver鍗氬＋涓撹瑙嗛銆傜壒鍒嚧璋ature鍜孌eepMind鎻愪緵璁伅鍜岃祫鏂欐巿鏉冦€侟/div>
銆€銆€Nature浠婂ぉ涓婄嚎鐨勮繖绡囬噸纾呰鏂囷紝璇︾粏浠嬬粛浜嗚胺姝孌eepMind鍥㈤槦鏈€鏂扮殑鐮旂┒鎴愭灉銆備汉宸ユ櫤鑳界殑涓€椤归噸瑕佺洰鏍囷紝鏄湪娌℃湁浠讳綍鍏堥獙鐭ヨ瘑鐨勫墠鎻愪笅锛岄€氳繃瀹屽叏鐨勮嚜瀛︼紝鍦ㄦ瀬鍏锋寫鎴樼殑棰嗗煙锛岃揪鍒拌秴浜虹殑澧冨湴銆傚幓骞达紝闃挎硶鐙楋紙AlphaGo锛変唬琛ㄤ汉宸ユ櫤鑳藉湪鍥存棰嗗煙棣栨鎴樿儨浜嗕汉绫荤殑涓栫晫鍐犲啗锛屼絾鍏舵鑹虹殑绮捐繘锛屾槸寤虹珛鍦ㄨ绠楁満閫氳繃娴烽噺鐨勫巻鍙叉璋卞涔犲弬鎮熶汉绫绘鑹虹殑鍩虹涔嬩笂锛岃繘鑰岃嚜鎴戣缁冿紝瀹炵幇瓒呰秺銆侟/div>
銆€銆€
闃垮皵娉曞厓妫嬪姏鐨勫闀夸笌绉垎姣擖/div>
銆€銆€鍙槸浠婂ぉ锛屾垜浠彂鐜帮紝浜虹被鍏跺疄鎶婇樋娉曠嫍鏁欏潖浜嗭紒鏂颁竴浠ｇ殑闃挎硶鍏?AlphaGo Zero),瀹屽叏浠庨浂寮€濮嬶紝涓嶉渶瑕佷换浣曞巻鍙叉璋辩殑鎸囧紩锛屾洿涓嶉渶瑕佸弬鑰冧汉绫讳换浣曠殑鍏堥獙鐭ヨ瘑锛屽畬鍏ㄩ潬鑷繁涓€涓汉寮哄寲瀛︿範锛坮einforcement learning锛夊拰鍙傛偀,妫嬭壓澧為暱杩滆秴闃挎硶鐙楋紝鐧炬垬鐧捐儨锛屽嚮婧冮樋娉曠嫍100-0銆侟/div>
銆€銆€杈惧埌杩欐牱涓€涓按鍑嗭紝闃挎硶鍏冨彧闇€瑕佸湪4涓猅PU涓婏紝鑺变笁澶╂椂闂达紝鑷繁宸﹀彸浜掓悘490涓囨灞€銆傝€屽畠鐨勫摜鍝ラ樋娉曠嫍锛岄渶瑕佸湪48涓猅PU涓婏紝鑺卞嚑涓湀鐨勬椂闂达紝瀛︿範涓夊崈涓囨灞€锛屾墠鎵撹触浜虹被銆侟/div>
銆€銆€杩欑瘒璁烘枃鐨勭涓€鍜岄€氳浣滆€呮槸DeepMind鐨凞avid Silver鍗氬＋,闃挎硶鐙楅」鐩礋璐ｄ汉銆備粬浠嬬粛璇撮樋娉曞厓杩滄瘮闃挎硶鐙楀己澶э紝鍥犱负瀹冧笉鍐嶈浜虹被璁ょ煡鎵€灞€闄愶紝鑰岃兘澶熷彂鐜版柊鐭ヨ瘑锛屽彂灞曟柊绛栫暐锛欬/div>
銆€銆€This technique is more powerful than previous versions of AlphaGo because it is no longer constrained by the limits of human knowledge.Instead,it is able to learn tabula rasa from the strongest player in the world:AlphaGo itself.AlphaGo Zero also discovered new knowledge,developing unconventional strategies and creative new moves that echoed and surpassed the novel techniques it played in the games against Lee Sedol and Ke Jie.
銆€銆€DeepMind鑱斿悎鍒涘浜哄拰CEO鍒欒杩欎竴鏂版妧鏈兘澶熺敤浜庤В鍐宠濡傝泲鐧借川鎶樺彔鍜屾柊鏉愭枡寮€鍙戣繖鏍风殑閲嶈闂锛欬/div>
銆€銆€AlphaGo Zero is now the strongest version of our program and shows how much progress we can make even with less computing power and zero use of human data.Ultimately we want to harness algorithmic breakthroughs like this to help solve all sorts of pressing real world problems like protein folding or designing new materials.
銆€銆€缇庡浗鐨勪袱浣嶆鎵嬪湪Nature瀵归樋娉曞厓鐨勬灞€鍋氫簡鐐硅瘎锛氬畠鐨勫紑灞€鍜屾敹瀹樺拰涓撲笟妫嬫墜鐨勪笅娉曞苟鏃犲尯鍒紝浜虹被鍑犲崈骞寸殑鏅烘収缁撴櫠锛岀湅璧锋潵骞堕潪鍏ㄩ敊銆備絾鏄腑鐩樼湅璧锋潵鍒欓潪甯歌寮傦細
銆€銆€the AI鈥檚 open卢ing choices and end-game methods have converged on ours鈥攕eeing it arrive at our sequences from first principles suggests that we haven鈥檛 been on entirely the wrong track.By contrast,some of its middle-game judgements are truly mysterious.
銆€銆€涓烘洿娣卞叆浜嗚В闃挎硶鍏冪殑鎶€鏈粏鑺傦紝鐭ョぞ閲囪浜嗙編鍥芥潨鍏嬪ぇ瀛︿汉宸ユ櫤鑳戒笓瀹堕檲鎬＄劧鏁欐巿銆備粬鍚戠煡绀句粙缁嶈锛欬/div>
銆€銆€DeepMind鏈€鏂版帹鍑虹殑AlphaGo Zero闄嶄綆浜嗚缁冨鏉傚害锛屾憜鑴变簡瀵逛汉绫绘爣娉ㄦ牱鏈?浜虹被鍘嗗彶妫嬪眬)鐨勪緷璧栵紝璁╂繁搴﹀涔犵敤浜庡鏉傚喅绛栨洿鍔犳柟渚垮彲琛屻€傛垜涓汉瑙夊緱鏈€鏈夎叮鐨勬槸璇佹槑浜嗕汉绫荤粡楠岀敱浜庢牱鏈┖闂村ぇ灏忕殑闄愬埗锛屽線寰€閮芥敹鏁涗簬灞€閮ㄦ渶浼樿€屼笉鑷煡锛堟垨鏃犳硶鍙戠幇锛夛紝鑰屾満鍣ㄥ涔犲彲浠ョ獊鐮磋繖涓檺鍒躲€備箣鍓嶅ぇ瀹堕殣闅愮害绾﹁寰楀簲璇ュ姝わ紝鑰岀幇鍦ㄦ槸閾佺殑閲忓寲浜嬪疄鎽嗗湪闈㈠墠锛?/div>
銆€銆€浠栬繘涓€姝ヨВ閲婇亾锛欬/div>
銆€銆€杩欑瘒璁烘枃鏁版嵁鏄剧ず瀛︿範浜虹被閫夋墜鐨勪笅娉曡櫧鐒惰兘鍦ㄨ缁冧箣鍒濊幏寰楄緝濂界殑妫嬪姏锛屼絾鍦ㄨ缁冨悗鏈熸墍鑳借揪鍒扮殑妫嬪姏鍗村彧鑳戒笌鍘熺増鐨凙lphaGo鐩歌繎锛岃€屼笉瀛︿範浜虹被涓嬫硶鐨凙lphaGo Zero鏈€缁堝嵈鑳借〃鐜板緱鏇村ソ銆傝繖鎴栬璇存槑浜虹被鐨勪笅妫嬫暟鎹皢绠楁硶瀵煎悜浜嗗眬閮ㄦ渶浼?local optima)锛岃€屽疄闄呮洿浼樻垨鑰呮渶浼樼殑涓嬫硶涓庝汉绫荤殑涓嬫硶瀛樺湪涓€浜涙湰璐ㄧ殑涓嶅悓锛屼汉绫诲疄闄呪€欒瀵尖€欎簡AlphaGo銆傛湁瓒ｇ殑鏄鏋淎lphaGo Zero鏀惧純瀛︿範浜虹被鑰屼娇鐢ㄥ畬鍏ㄩ殢鏈虹殑鍒濆涓嬫硶锛岃缁冭繃绋嬩篃涓€鐩存湞鐫€鏀舵暃鐨勬柟鍚戣繘琛岋紝鑰屾病鏈変骇鐢熼毦浠ユ敹鏁涚殑鐜拌薄銆侟/div>
銆€銆€闃挎硶鍏冩槸濡備綍瀹炵幇鏃犲笀鑷€氱殑鍛紵鏉滃厠澶у鍗氬＋鐮旂┒鐢熷惔鏄ラ箯鍚戠煡绀句粙缁嶄簡鎶€鏈粏鑺傦細
銆€銆€涔嬪墠鎴樿儨鏉庝笘鐭崇殑AlphaGo鍩烘湰閲囩敤浜嗕紶缁熷寮哄涔犳妧鏈啀鍔犱笂娣卞害绁炵粡缃戠粶DNN瀹屾垚鎼缓锛岃€孉lphaGo Zero鍚稿彇浜嗘渶鏂版垚鏋滃仛鍑轰簡閲嶅ぇ鏀硅繘銆侟/div>
銆€銆€棣栧厛锛屽湪AlphaGo Zero鍑虹幇涔嬪墠锛屽熀浜庢繁搴﹀涔犵殑澧炲己瀛︿範鏂规硶鎸夌収浣跨敤鐨勭綉缁滄ā鍨嬫暟閲忓彲浠ュ垎涓轰袱绫胡涓€绫讳娇鐢ㄤ竴涓狣NN"绔埌绔?鍦板畬鎴愬叏閮ㄥ喅绛栬繃绋?姣斿DQN)锛岃繖绫绘柟娉曟瘮杈冭交渚匡紝瀵逛簬绂绘暎鍔ㄤ綔鍐崇瓥鏇撮€傜敤;鍙︿竴绫讳娇鐢ㄥ涓狣NN鍒嗗埆瀛︿範policy鍜寁alue绛?姣斿涔嬪墠鎴樿儨鏉庝笘鐭崇殑AlphaGoGo)锛岃繖绫绘柟娉曟瘮杈冨鏉傦紝瀵逛簬鍚勭鍐崇瓥鏇撮€氱敤銆傛娆＄殑AlphaGo Zero缁煎悎浜嗕簩鑰呴暱澶勶紝閲囩敤绫讳技DQN鐨勪竴涓狣NN缃戠粶瀹炵幇鍐崇瓥杩囩▼锛屽苟鍒╃敤杩欎釜DNN寰楀埌涓ょ杈撳嚭policy鍜寁alue锛岀劧鍚庡埄鐢ㄤ竴涓挋鐗瑰崱缃楁悳绱㈡爲瀹屾垚褰撳墠姝ラ閫夋嫨銆侟/div>
銆€銆€鍏舵锛孉lphaGo Zero娌℃湁鍐嶅埄鐢ㄤ汉绫诲巻鍙叉灞€锛岃缁冭繃绋嬩粠瀹屽叏闅忔満寮€濮嬨€傞殢鐫€杩戝嚑骞存繁搴﹀涔犵爺绌跺拰搴旂敤鐨勬繁鍏ワ紝DNN鐨勪竴涓己鐐规棩鐩婃槑鏄晋璁粌杩囩▼闇€瑕佹秷鑰楀ぇ閲忎汉绫绘爣娉ㄦ牱鏈紝鑰岃繖瀵逛簬灏忔牱鏈簲鐢ㄩ鍩?姣斿鍖荤枟鍥惧儚澶勭悊)鏄笉鍙兘鍔炲埌鐨勩€傛墍浠ew-shot learning鍜孴ransfer learning绛夊噺灏戞牱鏈拰浜虹被鏍囨敞鐨勬柟娉曞緱鍒版櫘閬嶉噸瑙嗐€侫lphaGo Zero鏄湪鍙屾柟鍗氬紙璁粌杩囩▼涓皾璇曡В鍐冲浜虹被鏍囨敞鏍锋湰鐨勪緷璧栵紝杩欐槸浠ュ線娌℃湁鐨勩€侟/div>
銆€銆€绗笁锛孉lphaGo Zero鍦―NN缃戠粶缁撴瀯涓婂惛鏀朵簡鏈€鏂拌繘灞曪紝閲囩敤浜哛esNet缃戠粶涓殑Residual缁撴瀯浣滀负鍩虹妯″潡銆傝繎鍑犲勾娴佽鐨凴esNet鍔犲ぇ浜嗙綉缁滄繁搴︼紝鑰孏oogLeNet鍔犲ぇ浜嗙綉缁滃搴︺€備箣鍓嶅ぇ閲忚鏂囪〃鏄庯紝ResNet浣跨敤鐨凴esidual缁撴瀯姣擥oogLeNet浣跨敤鐨処nception缁撴瀯鍦ㄨ揪鍒扮浉鍚岄娴嬬簿搴︽潯浠朵笅鐨勮繍琛岄€熷害鏇村揩銆侫lphaGo Zero閲囩敤浜哛esidual搴旇鏈夐€熷害鏂归潰鐨勮€冭檻銆侟/div>
銆€銆€鏉滃厠澶у鍗氬＋鐮旂┒鐢熻阿鐭ラ仴瀵规鍋氫簡杩涗竴姝ラ槓杩帮細
銆€銆€DeepMind鐨勬柊绠楁硶AlphaGo Zero寮€濮嬫憜鑴卞浜虹被鐭ヨ瘑鐨勪緷璧栵細鍦ㄥ涔犲紑濮嬮樁娈垫棤闇€鍏堝涔犱汉绫婚€夋墜鐨勮蛋娉曪紝鍙﹀杈撳叆涓病鏈変簡浜哄伐鎻愬彇鐨勭壒寰併€侟/div>
銆€銆€鍦ㄧ綉缁滅粨鏋勭殑璁捐涓婏紝鏂扮殑绠楁硶涓庝箣鍓嶇殑AlphaGo鏈変袱涓ぇ鐨勫尯鍒€傞鍏堬紝涓庝箣鍓嶅皢璧板瓙绛栫暐(policy)缃戠粶鍜岃儨鐜囧€?value)缃戠粶鍒嗗紑璁粌涓嶅悓锛屾柊鐨勭綉缁滅粨鏋勫彲浠ュ悓鏃惰緭鍑鸿姝ョ殑璧板瓙绛栫暐(policy)鍜屽綋鍓嶆儏褰笅鐨勮儨鐜囧€?value)銆傚疄闄呬笂policy涓巚alue缃戠粶鐩稿綋浜庡叡鐢ㄤ簡涔嬪墠澶ч儴鍒嗙殑鐗瑰緛鎻愬彇灞傦紝杈撳嚭闃舵鐨勬渶鍚庡嚑灞傜粨鏋勪粛鐒舵槸鐩镐簰鐙珛鐨勩€傝缁冪殑鎹熷け鍑芥暟涔熷悓鏃跺寘鍚簡policy鍜寁alue涓ら儴鍒嗐€傝繖鏍风殑鏄剧劧鑳藉鑺傜渷璁粌鏃堕棿锛屾洿閲嶈鐨勬槸娣峰悎鐨刾olicy涓巚alue缃戠粶涔熻鑳介€傚簲鏇村绉嶄笉鍚屾儏鍐点€侟/div>
銆€銆€鍙﹀涓€涓ぇ鐨勫尯鍒湪浜庣壒寰佹彁鍙栧眰閲囩敤浜?0鎴?0涓畫宸ā鍧楋紝姣忎釜妯″潡鍖呭惈2涓嵎绉眰銆備笌涔嬪墠閲囩敤鐨?2灞傚乏鍙崇殑鍗风Н灞傜浉姣旓紝娈嬪樊妯″潡鐨勮繍鐢ㄤ娇缃戠粶娣卞害鑾峰緱浜嗗緢澶х殑鎻愬崌銆侫lphaGo Zero涓嶅啀闇€瑕佷汉宸ユ彁鍙栫殑鐗瑰緛搴旇涔熸槸鐢变簬鏇存繁鐨勭綉缁滆兘鏇存湁鏁堝湴鐩存帴浠庢鐩樹笂鎻愬彇鐗瑰緛銆傛牴鎹枃绔犳彁渚涚殑鏁版嵁锛岃繖涓ょ偣缁撴瀯涓婄殑鏀硅繘瀵规鍔涚殑鎻愬崌璐＄尞澶ц嚧鐩哥瓑銆侟/div>
銆€銆€鍥犱负杩欎簺鏀硅繘锛孉lphaGo Zero鐨勮〃鐜板拰璁粌鏁堢巼閮芥湁浜嗗緢澶х殑鎻愬崌锛屼粎閫氳繃4鍧桾PU鍜?2灏忔椂鐨勮缁冨氨鑳藉鑳滆繃涔嬪墠璁粌鐢ㄦ椂鍑犱釜鏈堢殑鍘熺増AlphaGo銆傚湪鏀惧純瀛︿範浜虹被妫嬫墜鐨勮蛋娉曚互鍙婁汉宸ユ彁鍙栫壒寰佷箣鍚庯紝绠楁硶鑳藉鍙栧緱鏇翠紭绉€鐨勮〃鐜帮紝杩欎綋鐜板嚭娣卞害绁炵粡缃戠粶寮哄ぇ鐨勭壒寰佹彁鍙栬兘鍔涗互鍙婂鎵炬洿浼樿В鐨勮兘鍔涖€傛洿閲嶈鐨勬槸锛岄€氳繃鎽嗚劚瀵逛汉绫荤粡楠屽拰杈呭姪鐨勪緷璧栵紝绫讳技鐨勬繁搴﹀己鍖栧涔犵畻娉曟垨璁歌兘鏇村鏄撳湴琚箍娉涘簲鐢ㄥ埌鍏朵粬浜虹被缂轰箯浜嗚В鎴栨槸缂轰箯澶ч噺鏍囨敞鏁版嵁鐨勯鍩熴€侟/div>
銆€銆€杩欎釜宸ヤ綔鎰忎箟浣曞湪鍛紵浜哄伐鏅鸿兘涓撳銆佺編鍥藉寳鍗＄綏鑾辩撼澶у澶忔礇鐗瑰垎鏍℃椽闊暀鎺堜篃瀵圭煡绀惧彂琛ㄤ簡鐪嬫硶锛欬/div>
銆€銆€鎴戦潪甯镐粩缁嗕粠澶村埌灏捐浜嗚繖绡囪鏂囥€傞鍏堣鑲畾宸ヤ綔鏈韩鐨勪环鍊笺€備粠鐢ㄦ璋?supervised learning)鍒版墧妫嬭氨锛屾槸閲嶅ぇ璐＄尞(contribution)锛佸共鎺変簡褰撳墠鏈€鐗涚殑妫嬫墜锛堝彉韬墠鐨勯樋娉曠嫍锛夛紝鏄痑dvancing state-of-the-art銆傜缁忕綉缁滅殑璁捐鍜岃缁冩柟娉曢兘鏈夋敼杩涳紝鏄垱鏂帮紙novelty锛夈€備粠搴旂敤瑙掑害锛屼互鍚庡彲鑳戒笉鍐嶉渶瑕佽€楄垂浜哄伐鍘讳负AI鐨勪骇鍝佸仛澶ч噺鐨勫墠鏈熷噯澶囧伐浣滐紝杩欐槸鍏舵剰涔?significance)鎵€鍦紒
銆€銆€鎺ョ潃锛屾椽鏁欐巿涔熺畝鍗曞洖椤句簡浜哄伐绁炵粡缃戠粶鐨勫巻鍙诧細
銆€銆€浜哄伐绁炵粡缃戠粶鍦ㄤ笂涓栫邯鍥涘崄骞翠唬灏卞嚭鏉ヤ簡锛屽皬鐏簡涓€涓嬪氨鎾戜笉涓嬪幓浜嗭紝鍏朵腑涓€涓師鍥犳槸澶у鍙戠幇杩欎笢瑗胯В鍐充笉浜嗏€滃紓鎴栭棶棰樷€濓紝鑰屼笖璁粌璧锋潵澶夯鐑︺€傚埌浜嗕笂涓栫邯涓冨崄骞翠唬锛孭aul Werbos璇诲崥鏃跺€欐嬁backpropagation鐨勭畻娉曟潵璁粌绁炵粡缃戠粶锛屾彁楂樹簡鏁堢巼锛岀敤澶氬眰绁炵粡缃戠粶鎶婂紓鎴栭棶棰樿В鍐充簡锛屼篃鎶婄缁忕綉缁滃甫鍏ヤ竴涓柊绾厓銆備笂涓栫邯鍏節鍗佸勾浠ｏ紝浜哄伐绁炵粡缃戠粶鐨勭爺绌惰繋鏉ヤ簡涓€鍦哄ぇ鐏紝瀛︽湳鍦堝彂浜嗘垚鍗冧笂涓囩瘒鍏充簬绁炵粡缃戠粶鐨勮鏂囷紝浠庤璁″埌璁粌鍒颁紭鍖栧啀鍒板悇琛屽悇涓氱殑搴旂敤銆侟/div>
銆€銆€Jim Burke鏁欐巿锛屼竴涓簲骞村墠閫€浼戠殑IEEE Life Fellow锛屾浘缁忚杩囬偅涓勾浠ｇ殑鏁呬簨锛氬幓寮€鐢靛姏绯荤粺鐨勫鏈細璁紝姣忚璁轰竴涓伐绋嬮棶棰橈紝涓嶇鏄暐锛屾€讳細鏈変竴甯汉璇磋繖鍙互鐢ㄧ缁忕綉缁滆В鍐筹紝褰撶劧鏈€鍚庝篃灏变笉浜嗕簡涔嬩簡銆傜畝鍗曠殑璇存槸澶у鎸栧潙鐏屾按鍚规场娉★紝鏈€鍚庢病鍟ュ彲蹇芥偁鐨勪簡锛屽氨鎵句釜鍒殑鍦板効鍐嶇户缁寲鍧戠亴姘村惞娉℃场銆備笂涓栫邯鏈殑瀛︽湳鍦堬紝濡傛灉鍑洪棬涓嶈鑷繁鎼炵缁忕綉缁滅殑閮戒笉濂芥剰鎬濊窡浜烘墦鎷涘懠锛屽氨鍜屽浠婄殑娣卞害瀛︿範銆佸ぇ鏁版嵁鍒嗘瀽涓€鏍枫€侟/div>
銆€銆€鐒跺悗锛屾椽鏁欐巿瀵逛汉宸ユ櫤鑳藉仛浜嗗苟涓嶅崄鍒嗕箰瑙傜殑灞曟湜锛欬/div>
銆€銆€鍥炲埌闃挎硶鐙椾笅妫嬭繖涓簨鍎匡紝浼撮殢鐫€澶ф暟鎹殑娴疆锛屾暟鎹寲鎺樸€佹満鍣ㄥ涔犮€佺缁忕綉缁滃拰浜哄伐鏅鸿兘绐佺劧闂村張鐏簡璧锋潵銆傝繖娆＄伀鐨勬湁娌℃湁鏂欏憿锛熸垜璁や负鏄湁鐨勶紝鏈夋捣閲忕殑鏁版嵁銆佹湁璁＄畻鑳藉姏鐨勬彁鍗囥€佹湁绠楁硶鐨勬敼杩涖€傝繖灏卞ソ姣斿綋骞存妸backpropagation鐢ㄥ湪绁炵粡缃戠粶涓婏紝鐨勭‘鏄釜绐佺牬銆侟/div>
銆€銆€鏈€缁堣繖涓伀鑳界儳澶氫箙锛岃繕寰楃湅绁炵粡缃戠粶鑳借В鍐冲灏戝疄闄呴棶棰樸€備簩鍗佸勾鍓嶇殑澶х伀涔嬪悗锛岃绁炵粡缃戠粶鈥滆В鍐斥€濈殑瀹為檯闂瀵ュ鏃犲嚑锛屽叾涓竴涓瘮杈冪煡鍚嶇殑鏄數鍔涜礋鑽烽娴嬮棶棰橈紝灏辨槸鐢ㄧ數閲忛娴嬶紝鍒氬ソ鏄垜鐨勪笓涓氥€傜敱浜庡綋骞寸缁忕綉缁滆繃浜庣伀鐖嗭紝瀵艰嚧绉戠爺閲嶅績鍑犱箮瀹屽叏绂诲紑浜嗕紶缁熺殑缁熻鏂规硶銆傜瓑鎴戝垰杩涘叆杩欎釜棰嗗煙鍋氬崥澹鏂囩殑鏃跺€欙紝灏辨嬁浼犵粺鐨勫鍏冨洖褰掓ā鍨嬬鏉€浜嗗競闈笂鐨勫悇绉嶇缁忕綉缁滈仐浼犵畻娉曠殑銆傛垜涓€璐殑鐪嬫硶锛屽浜庣溂鍓嶆祦琛岀殑涓滆タ锛屼笉瑕佺洸鐩拷閫愶紝瑕佸厛瀹℃椂搴﹀娍锛岀湅鐪嬭嚜宸辨搮闀垮暐銆佹湁鍟ョН绱紝鐪嬪噯浜嗗潙鍐嶈烦銆侟/div>
銆€銆€缇庡浗瀵嗘瓏鏍瑰ぇ瀛︿汉宸ユ櫤鑳藉疄楠屽涓讳换Satinder Singh涔熻〃杈句簡鍜屾椽鏁欐巿绫讳技鐨勮鐐癸細杩欏苟闈炰换浣曠粨鏉熺殑寮€濮嬶紝鍥犱负浜哄伐鏅鸿兘鍜屼汉鐢氳嚦鍔ㄧ墿鐩告瘮锛屾墍鐭ユ墍鑳戒緷鐒舵瀬绔湁闄愶細
銆€銆€This is not the beginning of any end because AlphaGo Zero,like all other successful AI so far,is extremely limited in what it knows and in what it can do compared with humans and even other animals.
銆€銆€涓嶈繃锛孲ingh鏁欐巿浠嶇劧瀵归樋娉曞厓澶у姞璧炶祻锛氳繖鏄竴椤归噸澶ф垚灏?鏄剧ず寮哄寲瀛︿範鑰屼笉渚濊禆浜虹殑缁忛獙锛屽彲浠ュ仛鐨勬洿濂斤細
銆€銆€The improvement in training time and computational complex卢ity of AlphaGo Zero relative to AlphaGo,achieved in about a year,is a major achieve卢ment鈥he results suggest that AIs based on reinforcement learning can perform much better than those that rely on human expertise.
銆€銆€闄堟€＄劧鏁欐巿鍒欏浜哄伐鏅鸿兘鐨勬湭鏉ュ仛浜嗚繘涓€姝ョ殑鎬濊€冿細
銆€銆€AlphaGo Zero娌℃湁浣跨敤浜虹被鏍囨敞锛屽彧闈犱汉绫荤粰瀹氱殑鍥存瑙勫垯锛屽氨鍙互鎺ㄦ紨鍑洪珮鏄庣殑璧版硶銆傛湁瓒ｇ殑鏄紝鎴戜滑杩樺湪璁烘枃涓湅鍒颁簡AlphaGo Zero鎺屾彙鍥存鐨勮繃绋嬨€傛瘮濡傚浣曢€愭笎瀛︿細涓€浜涘父瑙佺殑瀹氬紡涓庡紑灞€鏂规硶锛屽绗竴鎵嬬偣涓変笁銆傜浉淇¤繖涔熻兘瀵瑰洿妫嬬埍濂借€呯悊瑙lphaGo鐨勪笅妫嬮鏍兼湁鎵€鍚彂銆侟/div>
銆€銆€闄や簡鎶€鏈垱鏂颁箣澶栵紝AlphaGo Zero鍙堜竴娆″紩鍙戜簡涓€涓€煎緱鎵€鏈変汉宸ユ櫤鑳界爺绌惰€呮€濊€冪殑闂:鍦ㄦ湭鏉ュ彂灞曚腑锛屾垜浠┒绔熷簲璇ュ浣曠湅寰呬汉绫荤粡楠岀殑浣滅敤銆傚湪AlphaGo Zero鑷富瀛︿細鐨勮蛋娉曚腑锛屾湁涓€浜涗笌浜虹被璧版硶涓€鑷达紝鍖哄埆涓昏鍦ㄤ腑闂寸浉鎸侀樁娈点€侫lphaGo Zero宸茬粡鍙互缁欎汉绫诲綋鍥存鑰佸笀锛屾寚瀵间汉绫绘€濊€冧箣鍓嶆病瑙佽繃鐨勮蛋娉曪紝鑰屼笉鐢ㄥ畬鍏ㄦ嫎娉ヤ簬鍥存澶у笀鐨勭粡楠屻€備篃灏辨槸璇碅lphaGo Zero鍐嶆鎵撶牬浜嗕汉绫荤粡楠岀殑绁炵鎰燂紝璁╀汉鑴戜腑褰㈡垚鐨勭粡楠屼篃鏄彲浠ヨ鎺㈡祴鍜屽涔犵殑銆侟/div>
銆€銆€闄堟暀鎺堟渶鍚庝篃鎻愬嚭涓€涓湁瓒ｇ殑鍛介锛欬/div>
銆€銆€鏈潵鎴戜滑瑕侀潰瀵圭殑涓€涓寫鎴樺彲鑳藉氨鏄?鍦ㄤ竴浜涗笌鏃ュ父鐢熸椿鏈夊叧鐨勫喅绛栭棶棰樹笂锛屼汉绫荤粡楠屽拰鏈哄櫒缁忛獙鍚屾椂瀛樺湪锛岃€屾満鍣ㄧ粡楠屼笌浜虹被缁忛獙鏈夊緢澶у樊鍒紝鎴戜滑鍙堣濡備綍鍘婚€夋嫨鍜屽埄鐢ㄥ憿锛烖/div>
銆€銆€涓嶈繃David Silver瀵规骞朵笉鎷呭績锛岃€屽鏈潵鍏呮弧淇″績銆備粬鎸囧嚭锛欬/div>
銆€銆€If similar techniques can be applied to other structured problems,such as protein folding,reducing energy consumption or searching for revolutionary new materials,the resulting breakthroughs have the potential to positively impact society.
鐩稿叧鏂伴椈
瑗垮畨涓鍏夌數绮惧瘑宸ョ▼鏈夐檺鍏徃
鐢佃瘽锛?29-61252307锛堥攢鍞儴锛?nbsp; 029-61206810锛堜汉鍔涜祫婧愶級
鐢佃瘽锛?29-61206811锛堥」鐩儴锛堻br /> 鍦板潃锛氳タ瀹夊競楂樻柊鍖烘瘯鍘熶簩璺浗瀹舵柊涓€浠ｄ汉宸ユ櫤鑳戒骇涓氬洯B7鍙锋ゼ
鐢佃瘽锛?29-68590659
浼犵湡锛?29-68590516
鍦板潃锛氳タ瀹夊競饩兼柊鍖轰笀饧嬩簲璺?鍙 鐜颁唬浼佷笟涓技涓滃尯3鍙锋ゼ
闂?/i>
浠濅汉鏅鸿兘绉戞妧锛堟睙鑻忥級鏈夐檺鍏徃
2011-12-12 12:15:59
鐢佃瘽锛?515-88558952
鐢佃瘽锛?8629369387锛堝紶鍏堢敓锛堻br> 鍦板潃锛氱洂鍩庡競鐩愬崡楂樻柊鍖烘柊鍥矾76鍙稢-2骞↑/p>
闂?/i>
浠濅汉鏅鸿兘绉戞妧锛堟垚閮斤級鏈夐檺鍏徃
2011-12-12 12:15:59
鐢佃瘽锛?28-64330937
鐢佃瘽锛?5982898576锛堟潕鍏堢敓锛堻br> 鍦板潃锛氭垚閮藉競楂樻柊瑗垮尯鏂颁笟璺?鍙锋眹閮芥€婚儴鍥簩鏈?鏍?鍗曞厓
闂?/i>
婀栧崡浠濅汉鏅鸿兘绉戞妧鏈夐檺鍏徃
2011-12-12 12:15:59
鐢佃瘽锛?3910828126锛堟潕鍏堢敓锛堻br> 鍦板潃锛氭箹鍗楃渷甯稿痉甯傞珮鏂板尯涓仈澶ч亾鍥涙湡鏍囧噯鍖栧巶鎴夸竴妤稽/p>
鍙嬫儏閾炬帴锛欬a href="https://www.opt.ac.cn/">涓闄㈣タ瀹夊厜鏈烘墍 涓鍒涙槦 瑗垮畨浜ら€氬ぇ瀛?/a> 瑗垮畨鐢靛瓙绉戞妧澶у
瑗垮畨涓鍏夌數绮惧瘑宸ョ▼鏈夐檺鍏徃鐗堟潈鎵€鏈
SAF Coolest v1.3.1.2 璁剧疆闈㈡澘 XBWSS-AATQ-UZZVE-ZFW
鍥剧墖ALT淇℃伅锛 瑗垮畨涓鍏夌數绮惧瘑宸ョ▼鏈夐檺鍏徃
鏃犳暟鎹彁绀
Sorry锛屽綋鍓嶆爮鐩鍦ㄦ洿鏂颁腑锛屾暚璇锋湡寰?
鎮ㄥ彲浠ユ煡鐪嬪叾浠栨爮鐩垨杩斿洖 棣栭〉