M'ndandanda wazopezekamo[Bisani][Show]
Dziko likusintha mwachangu chifukwa cha luntha lochita kupanga, komanso kuphunzira pamakina, komwe kumakhudza mbali iliyonse ya moyo wathu watsiku ndi tsiku.
Kuchokera kwa othandizira mawu omwe amagwiritsa ntchito NLP ndi kuphunzira pamakina kusungitsa nthawi, kuyang'ana zochitika pa kalendala yathu, ndikuyimba nyimbo ku zida zomwe zili zolondola kwambiri kuti athe kuyembekezera zosowa zathu tisanaziganizire.
Makompyuta amatha kusewera chess, kuchita opaleshoni, ndikukula kukhala makina anzeru, ngati anthu mothandizidwa ndi makina ophunzirira makina.
Tili m'nthawi ya kupita patsogolo kwaukadaulo, ndipo powona momwe makompyuta apangidwira m'kupita kwanthawi, titha kulosera zomwe zidzachitike m'tsogolo.
Kukhazikitsa demokalase kwa zida zamakompyuta ndi njira ndi chimodzi mwazinthu zazikulu zakusinthaku komwe kumawonekera. Asayansi azidziwitso apanga makompyuta amphamvu osokoneza deta m'zaka zisanu zapitazi pogwiritsa ntchito njira zamakono. Zotsatira zake ndi zodabwitsa.
Mu positi iyi, tiyang'ana mozama makina kuphunzira ma algorithms ndi kusiyanasiyana kwawo konse.
Ndiye, ma algorithms a Machine Learning ndi chiyani?
Njira yogwiritsiridwa ntchito ndi dongosolo la AI kuti igwire ntchito yake-kawirikawiri, kulosera zamtengo wapatali kuchokera kuzomwe zaperekedwa-imadziwika kuti makina ophunzirira makina.
Makina ophunzirira makina ndi njira yomwe imagwiritsa ntchito deta ndipo imagwiritsidwa ntchito popanga makina ophunzirira omwe ali okonzeka kupangidwa. Ngati kuphunzira pamakina ndi sitima yomwe imagwira ntchito, ndiye kuti ma aligorivimu ophunzirira makina ndi ma locomotives omwe amayendetsa ntchitoyo.
Njira yabwino yophunzirira makina ogwiritsa ntchito imatsimikiziridwa ndi vuto labizinesi yomwe mukuyesera kuthana nayo, mtundu wa data yomwe mukugwiritsa ntchito, ndi zomwe muli nazo.
Ma algorithms ophunzirira makina ndi omwe amasintha deta kukhala chitsanzo. Kutengera ndi vuto lomwe mukuyesera kuyankha, mphamvu yosinthira yomwe ilipo, ndi mtundu wa data yomwe muli nayo, kuyang'aniridwa, osayang'aniridwa, kapena kulimbikitsa ma aligorivimu ophunzirira amatha kuchita bwino.
Kotero, tidakambirana za maphunziro oyang'aniridwa, osayang'aniridwa, ndi kulimbikitsa, koma ndi chiyani? Tiyeni tiwafufuze.
Maphunziro Oyang'aniridwa, Osayang'aniridwa & Kulimbikitsa
Kuphunzira Kuyang'aniridwa
Pakuphunzira koyang'aniridwa, chitsanzo cha AI chimapangidwa potengera zomwe zaperekedwa komanso chizindikiro chomwe chikuyimira zotsatira zomwe zanenedweratu. Kutengera zolowa ndi zotuluka, choyimiracho chimapanga equation yamapu, ndipo pogwiritsa ntchito equation yamapuyo, imaneneratu za zolowa m'tsogolomu.
Tiyerekeze kuti tifunika kupanga chitsanzo chomwe chingasiyanitse galu ndi mphaka. Zithunzi zambiri za amphaka ndi agalu zimadyetsedwa mu chitsanzocho ndi zilembo zosonyeza ngati ali amphaka kapena agalu kuti aphunzitse chitsanzocho.
Chitsanzocho chikufuna kukhazikitsa equation yokhudzana ndi zolemba pazithunzi zojambulidwa ndi zithunzizo. Ngakhale ngati chitsanzocho sichinawonepo chithunzicho, pambuyo pa maphunziro, chikhoza kuzindikira ngati ndi mphaka kapena galu.
Kuphunzira Osayang'aniridwa
Kuphunzira kosayang'aniridwa kumaphatikizapo kuphunzitsa mtundu wa AI pazowonjezera popanda kuzilemba. Chitsanzochi chimagawaniza zomwe zalowetsazo m'magulu omwe ali ndi makhalidwe ofanana.
Zolemba zam'tsogolo zomwe zalowetsazo zimanenedweratu kutengera momwe mawonekedwe ake akufananira ndi gulu limodzi. Ganizirani momwe tiyenera kugawaniza gulu la mipira yofiira ndi yabuluu m'magulu awiri.
Tiyerekeze kuti mawonekedwe ena a mipira ndi ofanana, kupatula mtundu. Pamaziko a momwe angagawire mipira m'magulu awiri, chitsanzocho chimayang'ana makhalidwe omwe ali osiyana pakati pa mipira.
Magulu aŵiri a mipira—mmodzi wabuluu ndi wina wofiira—amapangidwa pamene mipirayo yagawidwa m’magulu aŵiri malinga ndi mtundu wake.
Maphunziro Owonjezera
Polimbikitsa kuphunzira, mtundu wa AI umafuna kukulitsa phindu lonse pochita monga momwe zingathere pazochitika zinazake. Ndemanga pa zotsatira zake zam'mbuyo zimathandiza chitsanzo kuphunzira.
Ganizirani za zochitika pamene loboti ikulangizidwa kuti isankhe njira pakati pa mfundo A ndi B. Loboti poyamba imasankha maphunziro aliwonse chifukwa alibe chidziwitso choyambirira.
Roboti imalandira zolowera panjira yomwe imatenga ndikumapeza chidziwitso kuchokera pamenepo. Loboti imatha kugwiritsa ntchito zolowetsayo kukonza vuto nthawi ina ikakumana ndi vuto lofananalo.
Mwachitsanzo, ngati loboti isankha njira B ndikulandila mphotho, monga mayankho abwino, imamvetsetsa nthawi ino kuti iyenera kusankha njira B kuti iwonjezere mphotho yake.
Tsopano potsiriza zomwe inu nonse mukuyembekezera, ndi ma aligorivimu.
Ma Algorithms Akuluakulu Ophunzirira Makina
1. Linear Regression
Njira yosavuta yophunzirira pamakina yomwe imapatuka pakuphunzira koyang'aniridwa ndikubwerera m'mizere. Ndi chidziwitso chochokera kumitundu yodziyimira payokha, chimagwiritsidwa ntchito kwambiri kuthetsa mavuto obwerera m'mbuyo ndikupanga kulosera pamitundu yodalira mosalekeza.
Kupeza mzere wokwanira bwino, womwe ungathandize kulosera zotsatira zamitundu yodalira mosalekeza, ndicho cholinga cha kutsika kwa mzere. Mitengo ya nyumba, zaka, ndi malipiro ndi zitsanzo za zinthu zomwe zimapitirirabe.
Chitsanzo chodziwika kuti kusinthasintha kwa mzere wosavuta amagwiritsa ntchito mzere wowongoka kuti awerengere mgwirizano pakati pa mtundu umodzi wodziimira ndi umodzi wodalira. Pali mitundu yopitilira iwiri yodziyimira payokha pamakwerero angapo.
Linear regression model ili ndi malingaliro anayi:
- Linearity: Pali kulumikizana kwa mzere pakati pa X ndi tanthauzo la Y.
- Homoscedasticity: Pa mtengo uliwonse wa X, kusiyana kotsalira kumakhala kofanana.
- Kudziyimira pawokha: Zowonera sizimayenderana ndi ufulu wodziyimira pawokha.
- Chizolowezi: X ikakhazikitsidwa, Y nthawi zambiri imagawidwa.
Kutsika kwa mizere kumachita modabwitsa pa data yomwe ingasiyanitsidwe motsatira mizere. Itha kuwongolera kuchulukitsitsa pogwiritsa ntchito njira zokhazikika, zotsimikizira, komanso njira zochepetsera milingo. Komabe, pali zochitika zina pomwe uinjiniya wokulirapo umafunika, zomwe nthawi zina zimatha kupangitsa kuti pakhale phokoso komanso phokoso.
2. Kubwerera kwa Logistic
Logistic regression ndi njira ina yophunzirira makina yomwe imachoka pakuphunzira koyang'aniridwa. Kugwiritsa ntchito kwake kwakukulu ndikugawika, pomwe kumatha kugwiritsidwa ntchito pamavuto obwerera.
Logistic regression imagwiritsidwa ntchito kulosera zamitundu yodalira pagulu pogwiritsa ntchito chidziwitso chochokera kuzinthu zodziyimira pawokha. Cholinga ndikugawa zotuluka, zomwe zitha kugwera pakati pa 0 ndi 1.
Kuchuluka kwa zolowetsa kumakonzedwa ndi sigmoid ntchito, ntchito yoyambitsa yomwe imasintha pakati pa 0 ndi 1.
Maziko obwerera m'mbuyo ndi kuyerekezera kwa kuthekera kwakukulu, njira yowerengetsera magawo a kugawanika koganiziridwa kupatsidwa deta yowonedwa.
3. Mtengo Wosankha
Njira ina yophunzirira makina yomwe imasiyanitsidwa ndi kuphunzira koyang'aniridwa ndi mtengo wosankha. Pazinthu zonse zamagulu ndi zochepetsera, njira yopangira chisankho ingagwiritsidwe ntchito.
Chida chopangira zisankhochi, chomwe chimafanana ndi mtengo, chimagwiritsa ntchito zowonetsera kuti ziwonetse zotsatira zomwe zikuyembekezeka, mtengo wake, ndi zotsatira zake. Pogawa deta m'magawo osiyana, lingalirolo ndi lofanana ndi malingaliro aumunthu.
Detayo yagawidwa m'magawo osiyanasiyana momwe tingathere. Cholinga chachikulu cha Mtengo wa Chisankho ndi kupanga njira yophunzitsira yomwe ingagwiritsidwe ntchito kulosera zamitundu yomwe mukufuna. Zomwe zikusowa zitha kuchitidwa zokha pogwiritsa ntchito Decision Tree.
Palibe chifukwa chojambulira chithunzi chimodzi, masinthidwe a dummy, kapena njira zina zopangiratu data. Ndizokhazikika m'lingaliro lakuti ndizovuta kuwonjezera deta yatsopano kwa izo. Ngati muli ndi chidziwitso chowonjezera, muyenera kukonzanso mtengowo pa dataset yonse.
Zotsatira zake, mitengo yachisankho ndi chisankho cholakwika pakugwiritsa ntchito kulikonse komwe kumafuna kusintha kosinthika kwachitsanzo.
Kutengera mtundu womwe mukufuna, mitengo yachisankho imagawidwa m'mitundu iwiri:
- Zosintha Zamagulu: Mtengo Wosankha momwe zosintha zamagulu zimakhalira.
- Zosiyanasiyana Zosasinthika: Mtengo Wosankha momwe cholinga chake chimakhala Chopitilira.
4. Nkhalango Yachisawawa
Random Forest Method ndiyo njira yotsatira yophunzirira makina ndipo ndi njira yoyang'anira yophunzirira makina yomwe imagwiritsidwa ntchito kwambiri pakugawira ndi kuyambiranso. Imakhalanso njira yopangira mtengo, yofanana ndi mtengo wosankha.
Nkhalango yamitengo, kapena mitengo yambiri yosankha, imagwiritsidwa ntchito mwachisawawa nkhalango kupanga ziweruzo. Pogwira ntchito zamagulu, njira yachisawawa ya nkhalango idagwiritsa ntchito zosinthika zamagulu pomwe ikugwira ntchito zobwerera m'mbuyo ndi ma dataset omwe amakhala ndi zosinthika mosalekeza.
Kuphatikizika, kapena kusanganikirana kwa mitundu yambiri, ndi zomwe njira yachisawawa ya nkhalango imachita, zomwe zikutanthauza kuti kulosera kumachitika pogwiritsa ntchito gulu lamitundu osati imodzi.
Kukhoza kugwiritsidwa ntchito pazovuta zonse zamagulu ndi zochepetsera, zomwe zimapanga makina ambiri amakono ophunzirira makina, ndi phindu lalikulu la nkhalango yosasinthika.
Njira ziwiri zosiyana zimagwiritsidwa ntchito ndi Ensemble:
- Kunyamula katundu: Pochita izi, zambiri zimapangidwa pagulu la maphunziro. Kuti muchepetse kusiyanasiyana kwa zolosera, izi zimachitika.
- Kukweza ndi njira yophatikizira ophunzira ofooka ndi ophunzira amphamvu pomanga zitsanzo zotsatizana, zomwe zimapangitsa kuti mtundu womaliza ukhale wolondola kwambiri.
5. Naive Bayes
Nkhani ya binary (awiri-awiri) komanso yamitundu yambiri imatha kuthetsedwa pogwiritsa ntchito njira ya Naive Bayes. Njira ikafotokozedwa pogwiritsa ntchito magawo a binary kapena gulu, ndiyosavuta kumvetsetsa. Lingaliro lopangidwa ndi gulu la Naive Bayes ndikuti kukhalapo kwa gawo limodzi mkalasi sikukhudzana ndi kupezeka kwa zina zilizonse.
Njira yomwe ili pamwambapa ikuwonetsa:
- P (H): Kuthekera kuti lingaliro H ndilolondola. Kuthekera koyambirira kumatchedwa izi.
- P (E): Kuthekera kwa umboni
- P (E | H): Mpata woti lingalirolo likuchirikizidwa ndi umboni.
- P(H|E): Mpata woti chiphunzitsocho ndi chowona, popereka umboni.
Kagulu ka Naive Bayes angaganizire chilichonse mwazochitazi payekhapayekha pozindikira kuthekera kwa zotsatira zina, ngakhale izi zitalumikizidwa. Mtundu wa Naive Bayesian ndiwosavuta kupanga komanso wogwira ntchito pamaseti akulu akulu.
Zimadziwika kuti zimagwira ntchito bwino kuposa ngakhale njira zovuta zogawira magulu pomwe ndizoyambira. Ndi gulu la ma algorithms omwe onse amatengera chiphunzitso cha Bayes, osati njira imodzi.
6. K-Oyandikana nawo Apafupi
Njira ya K-Nearest Nearest (kNN) ndi kagawo kakang'ono ka kuphunzira pamakina koyang'aniridwa komwe kungagwiritsidwe ntchito kuthana ndi zovuta zamagulu ndi kuyambiranso. Ma algorithm a KNN amalingalira kuti zinthu zofananira zitha kupezeka pafupi.
Ndimakumbukira ngati gulu la anthu amalingaliro ofanana. kNN imapindula ndi lingaliro la kufanana pakati pa ma data ena pogwiritsa ntchito kuyandikira, kuyandikira, kapena mtunda. Kuti mulembe zomwe sizikuwoneka potengera malo omwe ali pafupi omwe ali pafupi, njira yamasamu imagwiritsidwa ntchito kuti mudziwe kusiyana pakati pa mfundo zomwe zili pa graph.
Muyenera kudziwa mtunda pakati pa ma data kuti muzindikire malo ofananira apafupi. Miyezo ya mtunda monga mtunda wa Euclidean, mtunda wa Hamming, mtunda wa Manhattan, ndi mtunda wa Minkowski ungagwiritsidwe ntchito pa izi. K imadziwika kuti nambala yoyandikana nayo pafupi, ndipo nthawi zambiri imakhala nambala yosamvetseka.
KNN ingagwiritsidwe ntchito pamagulu ndi zovuta zochepetsera. Kuneneratu komwe kunachitika KNN ikagwiritsidwa ntchito kukonzanso zinthu kumatengera tanthauzo kapena pakati pa zochitika zofananira za K.
Zotsatira za kalembedwe kagulu kotengera KNN zitha kuzindikirika ngati kalasi yomwe ili ndi ma frequency apamwamba kwambiri pakati pa zochitika za K zofananira. Nthawi zonse amavotera gulu lawo, ndipo zoloserazo zimakhala za gulu lomwe limalandira mavoti ochulukirapo.
7. K-njira
Ndi njira yophunzirira mosayang'aniridwa yomwe imayang'anira zovuta zamagulu. Ma seti a data amagawidwa m'magulu angapo - tiyitane tiyeni tiyimbe K - m'njira yoti ma data a gulu lililonse akhale ofanana komanso osiyana ndi omwe ali m'magulu ena.
K-amatanthauza njira yophatikizira:
- Pagulu lililonse, algorithm ya K-means imasankha k centroids, kapena mfundo.
- Ndi ma centroids apafupi kapena magulu a K, mfundo iliyonse imapanga gulu.
- Tsopano, ma centroids atsopano amapangidwa kutengera mamembala amgulu omwe alipo kale.
- Mtunda wapafupi kwambiri wa mfundo iliyonse umawerengeredwa pogwiritsa ntchito ma centroid osinthidwawa. Mpaka ma centroids asasinthe, njirayi imabwerezedwa.
Ndi yachidule, yodalirika komanso yosavuta kuimvetsa. Ngati pali zovuta, kusinthasintha kwa k-means kumapangitsa kusintha kukhala kosavuta. Pamene magulu a data ali osiyana kapena olekanitsidwa bwino ndi mzake, zotsatira zake zimakhala zabwino kwambiri. Sichingathe kusamalira deta yolakwika kapena zotsatsa.
8. Makina Othandizira Vector
Mukamagwiritsa ntchito njira ya SVM kugawa deta, deta yaiwisi imawonetsedwa ngati madontho mu n-dimensional space (pomwe n ndi chiwerengero cha zomwe muli nazo). Zomwezo zitha kugawidwa mosavuta chifukwa mtengo wa chinthu chilichonse umalumikizidwa ndi dongosolo linalake.
Kuti mulekanitse deta ndikuyiyika pa graph, gwiritsani ntchito mizere yotchedwa classifiers. Njirayi imakonza mfundo iliyonse ya deta ngati mfundo mu malo a n-dimensional, pomwe n ndi chiwerengero cha zinthu zomwe muli nazo ndipo mtengo wamtundu uliwonse ndi mtengo wogwirizanitsa.
Tsopano tipeza mzere womwe umagawaniza deta m'maseti awiri a data omwe adagawidwa mosiyana. Mipata yochokera kumalo oyandikana nawo m'magulu awiriwa idzakhala yotalikirana kwambiri pamzerewu.
Popeza mfundo ziwiri zapafupi ndizo zomwe zili kutali kwambiri ndi mzere mu chitsanzo pamwambapa, mzere womwe umagawanitsa deta m'magulu awiri omwe adagawidwa mosiyana ndi mzere wapakati. Gulu lathu ndi mzere uwu.
9. Kuchepetsa Dimensionality
Pogwiritsa ntchito njira yochepetsera kukula, deta yophunzitsira ikhoza kukhala ndi zosintha zochepa. M'mawu osavuta, amatanthauza njira yochepetsera kukula kwa mawonekedwe anu. Tiyerekeze kuti deta yanu ili ndi mizati 100; kuchepa kwa dimensionality kudzachepetsa kuchuluka kwa magawo 20.
Mtunduwu umangokulirakulira ndipo umakhala ndi chiopsezo chokulirapo pomwe kuchuluka kwazinthu kumakwera. Vuto lalikulu kwambiri pakugwira ntchito ndi data mumiyeso yayikulu ndi yomwe imadziwika kuti "temberero la kukula," zomwe zimachitika data yanu ikachulukirachulukira.
Zinthu zotsatirazi zitha kugwiritsidwa ntchito kutsitsa dimensionality:
- Kuti mupeze ndi kusankha mawonekedwe oyenera, kusankha kwa mawonekedwe kumagwiritsidwa ntchito.
- Pogwiritsa ntchito zomwe zidalipo kale, uinjiniya wazinthu umapanga pamanja zatsopano.
Kutsiliza
Kuphunzira makina osayang'aniridwa kapena kuyang'aniridwa ndizotheka zonse. Sankhani maphunziro oyang'aniridwa ngati deta yanu ili yochepa komanso yodziwika bwino yophunzitsidwa.
Ma data akuluakulu nthawi zambiri amatha kuchita bwino ndikutulutsa zotsatira zabwino pogwiritsa ntchito maphunziro osayang'aniridwa. Kuphunzira kwambiri njira ndi zabwino kwambiri ngati muli ndi deta yochuluka yomwe imapezeka mosavuta.
Zolimbikitsa kuphunzira ndi kuphunzira kulimbikitsa mwakuya ndi mitu yomwe mwaphunzira. Makhalidwe, ntchito, ndi zopinga za Neural network tsopano zamveka kwa inu. Pomaliza, mudaganizira zosankha za zilankhulo zosiyanasiyana zamapulogalamu, ma IDE, ndi nsanja zikafika popanga zanu. mitundu yophunzirira makina.
Chotsatira chomwe muyenera kuchita ndikuyamba kuphunzira ndikugwiritsa ntchito chilichonse makina kuphunzira njira. Ngakhale nkhaniyo ili yotakata, mutu uliwonse ukhoza kumveka m’maola ochepa ngati muunika kuzama kwake. Mutu uliwonse umayima wokha kwa ena.
Muyenera kuganizira nkhani imodzi panthawi, kuiwerenga, kuigwiritsa ntchito, ndikugwiritsa ntchito chilankhulo chomwe mukufuna kuti mugwiritse ntchito ma aligorivimu mmenemo.
Siyani Mumakonda