Zviri Mukati[Viga][Ratidza]
Nyika iri kukurumidza kuchinja nekuda kwehungwaru hwekugadzira, uye kudzidza muchina, uko kune chekuita pane ese ehupenyu hwedu hwemazuva ese.
Kubva kune vabatsiri vezwi vanoshandisa NLP uye muchina kudzidza kubhuka nguva, tarisa zviitiko pakarenda yedu, uye ridza mimhanzi kune zvishandiso zvine chokwadi zvekuti vanogona kutarisira zvatinoda isu tisati tazvifunga.
Makomputa anogona kutamba chess, kuvhiya, uye kukura kuita akangwara, akawanda-akaita semichina nerubatsiro rwemuchina kudzidza algorithms.
Isu tiri panguva yekuenderera mberi kwetekinoroji, uye nekuona magadzirirwo akagadzirwa nemakomputa nekufamba kwenguva, tinogona kufanotaura nezve zvichaitika mune ramangwana.
Democratization yezvishandiso zvemakomputa uye nzira ndechimwe chezvinhu zvakakosha zveshanduko iyi inomira pachena. Data masayendisiti vakagadzira makomputa ane simba-anopwanya data mukati memakore mashanu apfuura nekuita zvisingaite nzira dzekucheka-kumucheto. Migumisiro yacho inoshamisa.
Mune ino post, isu tinotarisa zvakanyanya machine learning algorithms uye zvese zvakasiyana-siyana.
Saka, chii chiri Machine Kudzidza algorithms?
Maitiro anoshandiswa neAI sisitimu kuita basa rayo-kazhinji, kufanotaura kukosha kwekubuda kubva kune yakapihwa data rekuisa-inozivikanwa semuchina kudzidza algorithm.
Muchina wekudzidza algorithm inzira inoshandisa data uye inoshandiswa kugadzira modhi yekudzidza yemuchina yakagadzirira kugadzirwa. Kana muchina kudzidza chiri chitima chinoita basa, saka muchina kudzidza maalgorithms ndiwo malocomotives anofambisa basa.
Iyo yakanakisa muchina yekudzidza maitiro ekushandisa ichatemwa nedambudziko rebhizinesi rauri kuyedza kugadzirisa, mhando yedataset yauri kushandisa, uye zviwanikwa zvaunazvo.
Machine kudzidza algorithms ndeaya anoshandura data seti kuita modhi. Zvichienderana nerudzi rwedambudziko rauri kuedza kupindura, simba rekugadzirisa riripo, uye rudzi rwedata raunaro, rakatariswa, risingatarisirwe, kana kusimbisa maalgorithms ekudzidza anogona kuita zvakanaka.
Saka, takataura pamusoro pekutariswa, kusatariswa, uye kusimbisa kudzidza, asi chii? Ngatizviongorore.
Kutariswa, Kusina Kutariswa & Kusimbisa Kudzidza
Yakarongedzwa Kudzidza
Mukudzidza kunotariswa, iyo AI modhi inogadzirwa zvichibva pane yakapihwa uye iyo label inomiririra zvakafanotaurwa. Zvichienderana nezvinopinza nezvabuda, modhi inogadzira equation yemepu, uye ichishandisa iyo equation yemepu, inofanotaura zita rezvipo mune ramangwana.
Ngatitii tinoda kugadzira muenzaniso unogona kusiyanisa imbwa nekatsi. Mapikicha akawanda ekitsi nembwa anodyiswa mumuenzaniso ane mavara anoratidza kana dziri katsi kana imbwa kuitira kudzidzisa modhi.
Iyo modhi inotsvaga kumisikidza equation inoenderana nemavara pamifananidzo yekuisa kune iyo mifananidzo. Kunyange kana iyo modhi isati yamboona chifananidzo ichi, mushure mekudzidziswa, inogona kuziva kana iri yekati kana imbwa.
Kudzidzira Kusina Kutarisirwa
Kudzidza kusingatarisirwe kunosanganisira kudzidzisa modhi yeAI chete pane zvekushandisa pasina kuzvinyora. Iyo modhi inokamura data rekuisa mumapoka ane hukama hunhu.
Iyo remangwana label yeiyo yekuisa inobva yafanotaurwa zvichienderana nekuti hunhu hwayo hunoenderana sei nehumwe hwemhando. Funga nezvemamiriro ezvinhu apo tinofanira kupatsanura boka remabhora matsvuku nebhuruu mumapoka maviri.
Ngatitorei kuti mabhora 'mamwe maitiro akafanana, kunze kweruvara. Pamusoro pekuti inogona sei kuparadzanisa mabhora mumapoka maviri, muenzaniso unotarisa maitiro akasiyana pakati pemabhora.
Masumbu maviri emabhora - rimwe bhuruu uye rimwe dzvuku - anogadzirwa kana mabhora akakamurwa kuita mapoka maviri zvichienderana nerudzi rwavo.
Kusimbisa Kudzidza
Mukusimbisa kudzidza, iyo AI modhi inotsvaga kuwedzera purofiti yakazara nekuita sezvainogona mune imwe mamiriro ezvinhu. Mhinduro pamhedzisiro yayo yekutanga inobatsira modhi kudzidza.
Funga nezvemamiriro ezvinhu apo robhoti inorayirwa kusarudza nzira pakati pemapoinzi A uye B. Robhoti inotanga kusarudza imwe yezvidzidzo nokuti haina ruzivo rwekare.
Iyo robhoti inogamuchira kupinza munzira yainotora uye inowana ruzivo kubva mairi. Robhoti rinogona kushandisa kupinza kugadzirisa nyaya nguva inotevera parinosangana nemamiriro akafanana.
Semuenzaniso, kana robhoti ikasarudza sarudzo B uye inogamuchira mubairo, senge mhinduro yakanaka, inonzwisisa panguva ino kuti inofanirwa kusarudza nzira B kuti iwedzere mubairo wayo.
Zvino pakupedzisira izvo zvamuri kumirira mese, ndiyo algorithms.
Yakakura Machine Kudzidza Algorithms
1. Linear Regression
Iyo yakapfava muchina yekudzidza nzira inotsauka kubva painotariswa kudzidza ndeye mutsara regression. Neruzivo kubva kumhando dzakazvimirira, rinonyanya kushandiswa kugadzirisa nyaya dzekudzoreredza uye kugadzira fungidziro pane zvinoramba zvichitsamira zvakasiyana.
Kutsvaga mutsara wezvakanakisa kukwana, izvo zvinogona kubatsira mukufanotaura mhedzisiro yezvinoramba zvichitsamira zvakasiyana, ndicho chinangwa chekudzoreredza mutsara. Mitengo yedzimba, zera, uye mihoro mimwe mienzaniso yezvinoramba zvichikosha.
Muenzaniso unozivikanwa sekuti simple linear regression unoshandisa mutsara wakatwasuka kuverenga mubatanidzwa pakati pechinhu chimwe chakazvimirira uye chinotsamira chimwe chete. Pane zvinodarika zviviri zvakazvimiririra zvakasiyana mumitsetse yakawanda regression.
Iyo mutsara regression modhi ine fungidziro ina dzinodzika:
- Linearity: Pane hukama hwemutsara pakati peX uye zvinoreva Y.
- Homoscedasticity: Kukosha kwese kwe X, musiyano wasara wakafanana.
- Kuzvimirira: Kucherechedzwa kwakazvimirira kune imwe neimwe maererano nekuzvimirira.
- Normality: Kana X yagadziriswa, Y inowanzogoverwa.
Linear regression inoita zvinoyemurika kune data rinogona kupatsanurwa nemitsara. Iyo inogona kudzora kukwirisa nekushandisa nguva dzose, muchinjika-kusimbisa, uye dimensionality kuderedza maitiro. Nekudaro, pane zviitiko apo yakawedzera chimiro engineering inodiwa, izvo zvinogona nguva nenguva kuguma nekuwedzeredza uye ruzha.
2. Logistic Regression
Logistic regression imwe nzira yekudzidza yemuchina inosiya kubva pakudzidza kwakatariswa. Kunyanya kushandiswa kwayo kurongedza, nepo ichigona zvakare kushandiswa kumatambudziko ekudzoreredza.
Logistic regression inoshandiswa kufanotaura iyo categorical dependent variable uchishandisa ruzivo kubva kune zvakazvimirira zvinhu. Chinangwa ndechekuronga zvinobuda, izvo zvinongodonha pakati pe0 ne1.
Huwandu hwakayerwa hwezvipimo hunogadziriswa ne sigmoid basa, activation function inoshandura kukosha pakati pe0 ne1.
Hwaro hwekudzoreredza kwemaitiro ndeyekuyeresa kungangove fungidziro, nzira yekuverenga maparamita eanofungirwa mukana wekugovera yakapihwa data rakacherechedzwa.
3. Muti Wechisarudzo
Imwe nzira yekudzidza yemuchina inoparadzana kubva pakudzidza kwakatariswa muti wesarudzo. Kune ese ari maviri emhando uye kudzoreredza nyaya, sarudzo yemuti nzira inogona kushandiswa.
Ichi chishandiso chekuita sarudzo, chakafanana nemuti, chinoshandisa zvinomiririra zvinoonekwa kuratidza zviito zvinotarisirwa mhedzisiro, mutengo, uye zvinokonzeresa. Nekukamura data muzvikamu zvakasiyana, pfungwa yacho inofananidzwa nepfungwa dzemunhu.
Iyo data yakakamurwa kuita zvikamu zvakasiyana zvakanyanya sezvataigona kuita granulate. Chinangwa chikuru cheMuti weChisarudzo ndechekugadzira modhi yekudzidzisa inogona kushandiswa kufanotaura kirasi yemhando yechinangwa. Hunhu husipo hunogona kubatwa otomatiki uchishandisa Muti weChisarudzo.
Iko hakuna chinodiwa kune imwe-pfuti encoding, dummy zvinosiyana, kana mamwe matanho ekutanga data. Iyo yakaoma mupfungwa yekuti zvakaoma kuwedzera data nyowani kwairi. Kana iwe uine yekuwedzera yakanyorwa data, iwe unofanirwa kudzidzisazve muti pane yese dataset.
Nekuda kweizvozvo, miti yesarudzo isarudzo isina kunaka kune chero application inoda dynamic modhi shanduko.
Zvichienderana nerudzi rwechinangwa chakasiyana, miti yesarudzo inokamurwa mumhando mbiri:
- Categorical Variable: Muti Wechisarudzo umo chinangwa chinoshanduka chiri Categorical.
- Kuenderera mberi Kuchinja: Muti Wechisarudzo umo chinangwa chinoshanduka chinoenderera.
4. Random Sango
Iyo Random Sango Method ndiyo inotevera muchina kudzidza maitiro uye inotariswa muchina kudzidza algorithm inoshandiswa zvakanyanya muchikamu uye kudzoreredza nyaya. Iyo zvakare inzira-yakavakirwa pamuti, yakafanana nemuti wesarudzo.
Sango remiti, kana miti yakawanda yesarudzo, inoshandiswa nemaitiro esango kuita mitongo. Pakubata mabasa ekuronga, nzira yesango isingaite yakashandisa mutsauko wechikamu uchibata mabasa ekudzoreredza nema dataset ane zvinoramba zvichichinja.
Kubatanidzwa, kana kusanganiswa kwemhando dzakawanda, ndizvo zvinoitwa nenzira yesango, zvinoreva kuti kufanotaura kunoitwa pachishandiswa boka remhando kwete imwe chete.
Iko kugona kushandiswa kune ese ari maviri emapoka uye matambudziko ekudzoreredza, ayo anoumba mazhinji emazuva ano emuchina kudzidza masisitimu, ibhenefiti yakakosha yesango risingaite.
Matanho maviri akasiyana anoshandiswa neEnsemble:
- Bagging: Nekuita izvi, data rakawanda rinogadzirwa yedhata rekudzidzisa. Kuti kuderedze kusiyanisa mune zvakafanotaurwa, izvi zvinoitwa.
- Boosting inzira yekubatanidza vadzidzi vasina simba nevadzidzi vakasimba nekuvaka mamodheru anotevedzana, zvichikonzera iyo yekupedzisira modhi ine huchokwadi hwepamusoro.
5. Naive Bayes
Iyo bhinari (mbiri-kirasi) uye yakawanda-kirasi classification nyaya inogona kugadziriswa uchishandisa iyo Naive Bayes maitiro. Kana iyo nzira ichitsanangurwa uchishandisa mabhinari kana chikamu chekuisa kukosha, zviri nyore kubata. Fungidziro yakaitwa naNaive Bayes classifier ndeyekuti kuvepo kwechimwe chinhu mukirasi hakuna chekuita pakuvapo kwechimwe chinhu.
Formula iri pamusoro inoratidza:
- P (H): Mukana wekuti hypothesis H ndeyechokwadi. Mukana wekare unonzi uyu.
- P (E): Mikana yeuchapupu
- P (E | H): Mukana wekuti iyo hypothesis inotsigirwa nehumbowo.
- P (H | E): Mukana wekuti iyo hypothesis ndeyechokwadi, yakapihwa humbowo.
A Naive Bayes classifier angatarise imwe neimwe yeaya maitiro ega kana achiona mukana weimwe mhedzisiro, kunyangwe hunhu uhu hwakabatana. Iyo Naive Bayesian modhi iri nyore kuvaka uye inoshanda kune yakakura dataset.
Iyo inozivikanwa kuita zvirinani kupfuura nyangwe dzakaomesesa nzira dzekuisa pazvinenge zvichikosha. Iyo muunganidzwa wealgorithms ayo ese akavakirwa paBayes 'Theorem, pane imwe nzira.
6. K-Vavakidzani Vepedyo
Iyo K-yepedyo vavakidzani (kNN) maitiro ichikamu cheanotariswa muchina kudzidza chinogona kushandiswa kugadzirisa kupatsanura uye kudzoreredza nyaya. Iyo KNN algorithm inofungidzira kuti zvinhu zvakafanana zvinogona kuwanikwa padyo.
Ndinozvirangarira tiri boka revanhu vane pfungwa dzakafanana. kNN inoita mukana weiyo pfungwa yekufanana pakati pemamwe mapoinzi data uchishandisa kuswedera, kuswedera, kana chinhambwe. Kuitira kuisa chiratidzo che data risingaonekwe zvichibva paari pedyo akanyorwa anocherechedzwa mapoinzi edata, nzira yemasvomhu inoshandiswa kuona kupatsanurwa pakati pemapoinzi pagirafu.
Iwe unofanirwa kuona chinhambwe pakati pemapoinzi data kuitira kuti uone ari pedyo anofananidzwa nzvimbo. Zviyero zvechinhambwe senge Euclidean chinhambwe, Hamming chinhambwe, Manhattan chinhambwe, uye Minkowski chinhambwe chinogona kushandiswa pane izvi. Iyo K inozivikanwa senhamba yevavakidzani iri pedyo, uye inowanzova nhamba isinganzwisisike.
KNN inogona kuiswa kune kurongedza uye kudzoreredza matambudziko. Kufanotaura kunoitwa kana KNN ichishandiswa kudzoreredza nyaya kunobva pazvinoreva kana pakati peK-zvakawanda zvakafanana zvinoitika.
Mhedzisiro yemhando yealgorithm yakavakirwa paKNN inogona kutariswa sekirasi ine yakanyanya frequency pakati peK zvakanyanya kufanana kuitika. Chese chiitiko chinovhota kukirasi yavo, uye kufanotaura ndekweboka rinogamuchira mavhoti akawanda.
7. K-zvinoreva
Inyanzvi yekudzidza isina anotariswa inogadzirisa nyaya dzekuunganidza. Maseti edata akakamurwa kuita imwe nhamba yemasumbu - kufona ngatiitei K - nenzira yekuti mapoinzi edhata ega ega ane homogeneous uye akasiyana kubva kune mamwe masumbu.
K-zvinoreva nzira yekubatanidza:
- Kune rimwe nerimwe sumbu, iyo K-zvinoreva algorithm inosarudza k centroids, kana mapoinzi.
- Iine macentroids ari padyo kana K masumbu, poindi yega yega data inoumba sumbu.
- Ikozvino, macentroid matsva anogadzirwa zvichienderana nenhengo dzecluster dzatovepo.
- Chinhambwe chepedyo chepoindi yega yega data inoverengerwa uchishandisa aya akagadziridzwa centroids. Kusvikira macentroids asachinja, maitiro aya anodzokororwa.
Inokurumidza, yakavimbika, uye iri nyore kunzwisisa. Kana paine nyaya, k-zvinoreva 'kuchinjika kunoita kuti kugadzirisa kuve nyore. Kana ma dataset akasarudzika kana akasarudzika kubva kune mumwe, mhedzisiro ndiyo yakanaka. Haikwanise kubata data rakashata kana kunze.
8. Tsigira Vector Machines
Paunenge uchishandisa iyo SVM nzira yekuisa data, mbishi data inoratidzwa semadoti mune n-dimensional nzvimbo (apo n ndiyo nhamba yezvimiro zvaunazvo). Iyo data inogona kurongeka nyore nyore nekuti kukosha kwechimwe nechimwe chinhu chinobva chabatanidzwa kune yakatarwa kurongeka.
Kuparadzanisa data uye kuiisa pagirafu, shandisa mitsetse inozivikanwa se classifiers. Iyi nzira inoronga poindi yega yega data senge poindi mune n-dimensional nzvimbo, uko n ndiyo nhamba yezvimiro zvaunazvo uye kukosha kwechimwe nechimwe chinhu chakakosha kurongeka.
Isu tichawana zvino mutsara unokamura data kuita maviri seti yedata yakakamurwa zvakasiyana. Nhambwe kubva kunzvimbo dziri padyo mune chimwe nechimwe chezvikwata zviviri izvi zvichange zvakaparadzana zvakanyanya mumutsara uyu.
Sezvo mapoinzi maviri epedyo ari iwo ari kure zvakanyanya kubva kumutsara mumuenzaniso uri pamusoro, mutsara unokamura data mumapoka maviri akaiswa muzvikamu zvakasiyana mutsara wepakati. Yedu yemhando yemutsara uyu.
9. Dimensionality Reduction
Uchishandisa nzira yekudzikisa dimensionality, data yekudzidziswa inogona kunge iine mashoma ekuisa akasiyana. Nemashoko akareruka, zvinoreva maitiro ekudzikisira saizi yeseti yako. Ngatifungei kuti dhatabheti yako ine 100 columns; kuderedzwa kwe dimensionality kuchadzikisira huwandu ihwohwo kusvika pamakumi maviri makoramu.
Iyo modhi inongoerekana yakura zvakanyanya uye ine njodzi yakakura yekuwedzeredza sezvo huwandu hwezvinhu hunokwira. Nyaya hombe nekushanda nedata muhukuru hukuru ndiyo inozivikanwa se "kutuka kwehukuru," iyo inoitika kana data rako riine huwandu hwakawandisa hwehunhu.
Zvinhu zvinotevera zvinogona kushandiswa kuita kuderedza dimensionality:
- Kuti uwane uye usarudze maitiro akakodzera, sarudzo yemhando inoshandiswa.
- Uchishandisa zvinhu zvagara zviripo, chimiro cheinjiniya chinogadzira maficha matsva.
mhedziso
Kusatariswa kana kutariswa muchina kudzidza zvese zvinogoneka. Sarudza inotariswa kudzidza kana yako data iri shoma uye yakanyatso cherechedzwa kudzidziswa.
Maseti makuru edata aiwanzoita uye kuburitsa mhedzisiro iri nani vachishandisa kudzidza kusingatarisirwe. Kudzidza zvakadzika nzira dzakanakisa kana uine yakakura data yekuunganidza iyo inowanikwa nyore.
Yekusimbisa kudzidza uye kudzidza kusimbaradza kwakadzama ndezvimwe zvidzidzo zvawakadzidza. Neural network maitiro, mashandisiro, uye zvimhingamipinyi zvave kujeka kwauri. Chekupedzisira asi chisiri chidiki, wakafunga sarudzo dzemitauro yakasiyana-siyana yekuronga, maIDE, uye mapuratifomu kana zvasvika pakugadzira yako. michina yekudzidza mamodheru.
Chinhu chinotevera chaunofanira kuita kutanga kudzidza uye kushandisa imwe neimwe machine learning kusvika. Kunyange kana chidzidzo chacho chakafara, chero musoro wenyaya unogona kunzwisiswa mumaawa mashomanana kana ukaisa pfungwa pakudzika kwawo. Chidzidzo chimwe nechimwe chinomira chega kubva kune vamwe.
Iwe unofanirwa kufunga nezvenyaya imwe panguva, kuidzidza, kuishandisa, uye kushandisa mutauro wesarudzo yako kuita algorithm(s) mairi.
Leave a Reply