5.2.1 Galaxy Zoo

Galaxy Zoo ta haɗu da ƙoƙarin masu bayar da agajin da ba su da kwarewa don tsara kundin galaxia miliyan.

Cibiyar Zhang Zoo ta yi girma daga matsalar da Kevin Schawinski ke fuskanta, dalibi mai digiri a Astronomy a Jami'ar Oxford a 2007. Da yake sauƙaƙewa kadan, Schawinski yana sha'awar galaxies, kuma ana iya rarraba tauraron dan adam ta hanyar ilimin halittar jiki-elliptical ko spiral-and ta launin launi ko ja. A wannan lokacin, kyakkyawar hikima tsakanin masu nazarin sararin samaniya shine tauraron da ke ciki, irin su Milky Way, suna da launin shuɗi (suna nuna samari) da kuma galaxies na elliptical sun kasance ja (nuna tsofaffi). Schawinski ya yi shakkar irin wannan hikima. Ya yi tsammanin cewa yayin da wannan tsari zai iya zama gaskiya a gaba ɗaya, akwai yiwuwar ƙididdigar yawa, kuma ta hanyar nazarin ɗakin waɗannan galaxies marar amfani-wadanda basu dace da yanayin da ake tsammani ba-zai iya koyon wani abu game da tsarin da an halicci tauraron dan adam.

Sabili da haka, abin da Schawinski yake buƙatar don kawar da hikima ta al'ada shi ne babban ɓangaren samfurori da aka kirkiro; wato, tauraron da aka ƙayyade a matsayin kogi ko kuma elliptical. Matsalar, duk da haka, ita ce hanyoyin da aka tsara na algorithmic don tsarawa ba su da kyau sosai don amfani da su don binciken kimiyya; a wasu kalmomin, rarraba darajar galaxies shine, a wancan lokacin, matsala da ta wahala ga kwakwalwa. Sabili da haka, abin da ake buƙata shi ne babban adadin mutane . Schawinski ya dauki wannan matsala tare da sha'awar ɗaliban ɗalibai. A cikin wani marathon na kwanaki goma sha biyu, ya iya rarraba filaye 50,000. Duk da cewa galaxies 50,000 na iya zama kamar mai yawa, to amma kawai kusan kashi 5 cikin dari na nau'in galaxies kusan miliyan daya da aka hotunan a cikin binciken Sloan Digital Sky. Schawinski ya fahimci cewa yana buƙatar karin tsarin daidaitawa.

Abin farin, shi dai itace cewa aiki na kassa taurari ba ya bukatar m horo a ilmin taurari. za ka iya sanar da wani ya yi da shi m da sauri. A wasu kalmomin, kuma kõ dã kassa taurari ne mai aiki da yake da wuya ga kwakwalwa, shi ne kyawawan sauki ga mutane. Saboda haka, lokacin da zaune a cikin wani mashaya a Oxford, Schawinski da kuma 'yan'uwanmu falakin Chris Lintott mafarkin wani website inda masu sa kai za su rarraba images of taurari. Bayan 'yan watanni, Galaxy Zoo da aka haife.

A shafin yanar gizon Zoo, masu ba da agaji za su yi minti kaɗan na horo; misali, koyon bambanci tsakanin karfin da ke da karfin gaske (adadi 5.2). Bayan wannan horarwa, kowane mai ba da hidima ya wuce wani nau'i mai sauƙi mai sauƙi-daidai yana rarraba ɗakunan 11 na 15 tare da sanannun sanannun-sannan kuma za a fara samfurin galaxies wanda ba a sani ba ta hanyar binciken yanar gizo mai sauki (adadi 5.3). Tsarin daga mai ba da taimako ga likitan sama zai faru a cikin minti 10 kuma kawai yana buƙatar wucewa mafi ƙasƙanci, ƙaddarar sauƙi.

Figure 5.2: Misalan manyan nau'o'i biyu na tauraron dan adam: karkace da kuma elliptical. Ayyukan Zoo na Zoo sun yi amfani da masu aikin sa kai fiye da 100,000 don rarraba fiye da 900,000 hotuna. An sake bugawa ta hanyar izini daga http://www.GalaxyZoo.org da kuma binciken Sloan Digital Sky .

Figure 5.3: Gidan shigarwa inda aka bukaci masu sa ran su kayyade hoto guda. An sake buga shi ta izini daga Chris Lintott bisa ga hoto daga Sloan Digital Sky Survey .

Galaxy Zoo ta janyo hankalin masu aikin sa kai na farko bayan da aka gabatar da aikin a cikin wani labarin, kuma a cikin kimanin watanni shida aikin ya karu ya ƙunshi fiye da mutane 100,000 masana kimiyya, mutanen da suka halarci saboda suna jin dadin aikin kuma suna so su taimakawa wajen inganta astronomy. Tare, waɗannan masu sa kai 100,000 sun ba da gudummawa fiye da miliyan 40, tare da yawancin ƙididdigar da ke fitowa daga ƙananan ƙananan ƙungiyoyi (Lintott et al. 2008) .

Masu bincike waɗanda suke da kwarewar samun takardar shaidar masu karatun digiri na farko zasu iya zama masu shakka game da ingancin bayanai. Duk da yake wannan shakku ba daidai ba ne, Zoo Zoo ta nuna cewa lokacin da aka ba da gudummawar da aka ba da gudummawa a tsabtace, da ƙwararru, da kuma haɗuwa, za su iya samar da sakamako mai kyau (Lintott et al. 2008) . An muhimmanci zamba domin samun taro don ƙirƙirar sana'a-ingancin bayanan ne redundancy, cewa shi ne, ciwon wannan aiki yi da mutane daban-daban. A cikin Zoo Zoo, akwai kimanin fasalin 40 da galaxy; masu bincike da ke amfani da masu bincike na jami'a ba za su iya samun wannan juriya ba saboda haka za su kasance da damuwa da ingancin kowane ɗayan ɗayan. Abin da masu aikin sa kai basu samu a horarwa ba, sun kasance suna da tsararren aiki.

Koda yake tare da ƙaddarar jimla ta galaxy, duk da haka, hada haɗin gwargwadon aikin sa kai don samar da ƙidayar yarjejeniya ya zama maras kyau. Saboda kalubale irin wannan kalubalen da ke faruwa a mafi yawan ayyukan bincike na mutum, yana da amfani muyi la'akari da matakai uku waɗanda masu bincike na Zoo Zoo suka yi amfani da su wajen rarraba ra'ayoyinsu. Na farko, masu bincike "tsaftace" bayanan ta hanyar cire fassarar ajiyar kariya. Alal misali, mutanen da suka ba da labarin galaxy guda ɗaya-wani abu da zai faru idan suna ƙoƙari su yi amfani da sakamakon-idan duk an rarraba su. Wannan da sauran tsabtataccen tsaftacewa sun cire kimanin kashi 4 cikin dari na duka rarrabawa.

Abu na biyu, bayan tsaftacewa, masu bincike sun buƙaci cire matakan da suka dace a rarraba. Ta hanyar jerin binciken bincike na ban sha'awa da aka saka a cikin aikin asali - alal misali, nuna wasu masu sa ran galaxy a monochrome maimakon launi-masu binciken sun gano wasu abubuwan da suka fi son yin amfani da su, irin su nagartaccen tsari don rarraba galaxies masu nisa kamar yadda ake kira galaxies (Bamford et al. 2009) . Daidaitawa ga waɗannan ƙaddamarwa na musamman yana da mahimmanci saboda ladabi ba ta kawar da mahimmanci ta atomatik; Yana taimaka kawai wajen kawar da kuskuren bazuwar.

A ƙarshe, bayan da ake tuhuma, masu bincike sun buƙaci hanyar da za su haɗu da ƙayyadaddun mutum don samar da ƙididdigar ra'ayi. Hanyar da ta fi sauƙi don hada haɓakawa ga kowane galaxy zai kasance da zaɓin ƙayyadewa mafi yawan jama'a. Duk da haka, wannan matakan zai ba kowane nauyin aikin sa kai daidai, kuma masu bincike sun yi zaton cewa wasu masu aikin sa kai sun fi kwarewa fiye da sauran. Sabili da haka, masu bincike sun kirkiro wani tsari mai mahimmanci wanda yake kokarin ƙaddamar da mafi kyawun ɗalibai kuma ya ba su nauyi.

Sabili da haka, bayan aiwatar da matakai uku-tsabtatawa, ƙwaƙwalwa, da nauyin nauyi - ƙungiyar bincike ta Zoo Zoo ta sauya nauyin masu aikin sa kai miliyan 40 a cikin wani ɓangare na ƙaddamarwa na mutuntaka. A lokacin da aka kwatanta wadannan ragowar Zoo Zoo tare da ƙananan ƙananan ƙoƙari na masu bincike da suka hada da Schawinski wanda ya taimakawa wahayi zuwa Galaxy Zoo, akwai yarjejeniya mai ƙarfi. Saboda haka, masu ba da taimako, a cikin ƙididdiga, sun iya samar da ƙaddamarwa mai kyau da kuma matakin da masu bincike basu iya daidaita ba (Lintott et al. 2008) . A gaskiya ma, ta hanyar kasancewa da ɗan adam ga irin waɗannan tarurruka, Schawinski, Lintott, da sauransu sun nuna cewa kawai kimanin kashi 80 cikin dari na mahaukaci sun bi alamar da ake tsammani-launin blue da e-mail-da yawa takardun rubutu an rubuta game da wannan binciken (Fortson et al. 2011) .

Bisa ga wannan batu, yanzu zaka iya ganin yadda Galaxy Zoo ya bi girke-girke-girke-girke, wannan girke-girke wanda aka yi amfani da shi don yawancin ayyukan bincike na mutum. Na farko, babban matsala ya rabu cikin chunks. A wannan yanayin, matsalar da ke tattare da galaxies miliyan daya an raba shi cikin matsaloli miliyoyi na rarraba nau'in galaxy. Na gaba, ana amfani da aiki akan kowane kullun da kansa. A wannan yanayin, masu ba da agaji suna rarraba kowane nau'i nau'i nau'in nau'i nau'i nau'i nau'i nau'i nau'i ne. A ƙarshe, ana haɓaka sakamakon don samar da sakamakon yakamata. A wannan yanayin, haɗa mataki ya haɗa da tsabtatawa, ƙwaƙwalwa, da nauyin nauyi don samar da ƙayyadaddun ra'ayi ga kowanne galaxy. Kodayake yawancin ayyukan suna amfani da wannan girke-girke na gari, kowane mataki yana buƙatar daidaitawa akan matsalar da aka magance. Alal misali, a cikin aikin ƙididdiga na mutum wanda aka bayyana a kasa, za a bi da wannan girke-girke, amma amfani da hada matakai zai zama daban.

Ga kungiyar Zoo Zoo, wannan aikin farko shine kawai farkon. Nan da nan sun fahimci cewa kodayake sun iya rarraba kusa da galaxia miliyan, wannan sikelin bai isa ya yi aiki tare da bincike na sama na zamani ba, wanda zai iya samar da hotuna game da biliyan 10 (Kuminski et al. 2014) . Don karɓar karuwa daga dala miliyan 1 zuwa biliyan 10-wani nau'i mai nauyin 10,000 Zoo Zoo zai bukaci kimanin sau 10,000 more mahalarta. Kodayake adadin masu sa kai a yanar-gizon yana da girma, ba iyaka ba ne. Sabili da haka, masu bincike sun gano cewa idan sun ci gaba da samun bayanai mai yawa, an buƙaci sabuwar, har ma mafi daidaitawa, da buƙata.

Saboda haka, Manda Banerji yana aiki tare da Schawinski, Lintott, da sauran mambobi na kungiyar Zoo Zoo (2010) -Dajin da ke koyar da su don rarraba galaxies. Ƙari musamman, ta amfani da ƙaddamarwar ɗan adam da aka gina ta hanyar Zoo Zoo, Banerji ya gina samfurin ilmantarwa na na'ura wanda zai iya hango asali akan samfurin galaxy na mutum dangane da halaye na hoton. Idan wannan samfurin zai iya haifar da ƙaddamar da ɗan adam tare da daidaitattun daidaito, to, ana iya amfani da su masu bincike na Zoo Zoo don ƙayyade yawan tauraron dan adam.

Babban tushen Banerji da abokan aiki shine ainihin kama da fasaha da aka saba amfani dasu a cikin bincike na zamantakewa, kodayake irin wannan kamance bazai iya bayyana ba a kallon farko. Na farko, Banerji da abokan aiki sun canza kowane hoton a cikin jerin siffofin da suka taƙaita dukiyarsa. Alal misali, don hotunan tauraron dan adam, za'a iya samun siffofi guda uku: adadin blue a cikin hoton, da bambanci a cikin haske daga cikin pixels, da kuma ragowar wadanda ba a fara ba. Zabin zaɓi na ainihi yana da muhimmin ɓangare na matsalar, kuma yana buƙatar gwaninta na gine-gine. Wannan mataki na farko, wanda ake kira fasalin injiniya , yana haifar da matakan bayanai tare da jere guda ɗaya da siffar kuma uku ginshiƙai masu kwatanta wannan hoton. Bayar da matakan bayanai da kuma samfurin da ake so (misali, ko mutum ya samarda wannan hoton a matsayin galaxy), mai bincike yayi kirkiro na ilimin lissafi ko na'ura-misali, ƙwaƙwalwar haɗari-wanda ke nuna matsayin ɗan adam dangane da fasali na hoton. A ƙarshe, mai bincike yana amfani da sigogi a cikin wannan samfurin lissafi don samar da ƙaddarar sabbin lambobi (adadi 5.4). A cikin ilmantarwa na na'ura, wannan hanya-ta yin amfani da misalai da aka lakafta don ƙirƙirar samfurin da za a iya lakafta sababbin bayanai-ana kiranta koyon ilimin .

Figure 5.4: Bayanan da aka kwatanta yadda Banerji et al. (2010) yi amfani da ƙaddamar da Galaxy Zoo don horar da samfurin ilimin gwaji don yin jinsin galaxy. Hotunan hotuna sun tuba a cikin wani nau'i na fasali. A wannan misali mai sauƙi, akwai siffofi guda uku (yawan blue a cikin hoton, bambancin dake cikin haske daga cikin pixels, da kuma raguwa na pixels wadanda ba'aɗa). Bayan haka, don alamar hotuna, ana amfani da alamar Zoo na Zoo don horar da samfurin haɓaka na na'ura. A ƙarshe, ana amfani da na'ura don ƙididdige ƙaddamarwa ga sauran ƙaho. Ina kira wannan tsari na lissafin mutum don taimakawa, domin, maimakon samun mutane magance matsala, yana da mutane suna gina dataset wanda za'a iya amfani dashi don horar da kwamfuta don magance matsalar. Amfani da wannan tsarin tsarin kulawar ɗan adam wanda ya taimaka da kwamfutarka shine ya ba ka damar daukar nauyin bayanai da yawa ba tare da iyaka ba kawai. Hotunan hotuna da aka haifa ta hanyar izini daga Sloan Digital Sky Survey .

Hanyoyin da ke cikin Banerji da abokan aiki na na'ura sun fi rikitarwa fiye da wadanda ke cikin misali na wasa - misali, ta yi amfani da fasali irin su "de Vaucouleurs fit ratio" - kuma tsarinta bai zama tsinkaye ba, yana da hanyar sadarwa ta wucin gadi. Ta amfani da siffofinta, da tsarinta, da kuma yarjejeniyar da aka tsara na Galaxy Zoo, ta iya ƙirƙirar ma'aunin akan kowane fasali, sannan kuma amfani da waɗannan ma'auni don yin tsinkaya game da rarrabuwa na tauraron dan adam. Alal misali, bincikenta ya gano cewa hotuna da '' Vaucouleurs 'da' yan kasuwa '' ba su da yawa sun kasance mafi yawancin tauraron dan adam. Da aka ba waɗannan nauyin nauyi, ta iya yin hango nesa da samfurin galaxy tare da daidaitattun daidaito.

Ayyukan Banerji da abokan aiki sun juya cikin Zoo Zoo cikin abin da zan kira tsarin bincike na mutum . Hanyar da ta fi dacewa ta yi tunani game da waɗannan matasan tsarin shine cewa maimakon samun mutane su warware matsala, suna da mutane su gina dataset wanda za a iya amfani dasu don horar da kwamfuta don magance matsalar. Wani lokaci, horar da kwamfuta don magance matsalar zai iya buƙatar misalai da dama, kuma hanya ɗaya ta samar da adadin misalai da yawa shine haɗin kai. Amfani da wannan tsarin da aka taimaka ta kwamfuta shine cewa yana ba ka damar karɓar yawancin bayanai da yawa ba tare da amfani da ƙima ba. Alal misali, wani mai bincike wanda ke dauke da nau'in jinsin mahallin mutane zai iya gina samfuri mai faɗi wanda za'a iya amfani dasu don rarraba biliyan daya ko ma da tarin biliyan. Idan akwai lambobi masu yawa, to, wannan nau'in ƙwayar mutum-kwamfuta shine ainihin bayani kawai. Wannan ƙarancin iyaka ba shi da kyauta, duk da haka. Gina fasahar ilmantarwa da za ta iya haifar da halayen ɗan adam shine matsala mai wuya, amma sa'a akwai wasu litattafai masu kyau da aka ba da su ga wannan batu (Hastie, Tibshirani, and Friedman 2009; Murphy 2012; James et al. 2013) .

Galaxy Zoo ne mai kyau kwatanci nawa da yawa tsarin lissafin mutum ya fara. Na farko, mai bincike yayi ƙoƙarin yin aikin ta kanta ko tare da karamin ƙungiyar masu taimakawa na bincike (misali, Schawinski's first classification effort). Idan wannan hanyar ba ta da ƙarfinta, mai bincike zai iya motsawa zuwa tsarin aikin mutum tare da mahalarta mahalarta. Amma, ga wasu matakan bayanai, aikin kirki na mutum ba zai isa ba. A wannan batu, masu bincike suna buƙatar gina tsarin tsarin kwakwalwa na mutum wanda ake amfani dasu don horar da samfurin ilmantarwa na na'ura wanda za'a iya amfani dasu da yawancin bayanai.