Onjiniyela bokufunda abasebenza ngomshini njengamanje ochwepheshe obukhokhelwa kakhulu obukhokhelwa e-UK?
Ngokuya ngamazinga.fyi, iholo elijwayelekile cishe £ 100k– Iphakeme kunenjiniyela yesoftware, onjiniyela be-AI, kanye nososayensi wedatha.
Kepha akusona nje mayelana ne-paycheque.
Njengonjiniyela wokufunda umshini, uthola ukubhekana nezinkinga ezihehayo, uvivinya amathuluzi wokusika, futhi empeleni athinte umhlaba.
Ngingakutshela kusuka ekuhlangenwe nakho kokuqala kwesandla – kungenye yemisebenzi ejabulisa kakhulu futhi egcwalisa.
Ngakho-ke kulesi sihloko, ngizokunikeza umgwaqo ocacile futhi olula wokufunda umgwaqo ukuze ube ngunjiniyela wokufunda umshini, kanye nezinsizakusebenza ezinhle kakhulu.
Ake singene kukho!
Maths kanye nezibalo
Ngithe isikhathi nesikhathi futhi, kepha izibalo nezibalo zikude kakhulu izinto okufanele uzifunde uma ufuna umsebenzi wokufunda ngomshini noma kwimininingwane ephelele.
Ubuchwepheshe buza bahambe, bacabangele iBlockchain ne-AI, kepha izibalo zihlala ziyisimangaliso esiyisisekelo kuyo yonke le minyaka.
Ngenhlanhla, awudingi ukuba ubunye ubuhlakani beMaths ukuze usebenze ekufundeni komshini, ngingakuqinisekisa ngenhliziyo yonke lokhu kusuka kulwazi lokuqala.
Izinga le-Maths elidingekayo lilingana kakhulu ngezinto ozifundiswe ngeminyaka yakho yokugcina yesikole esiphakeme kanye nonyaka wokuqala noma ezimbili zeziqu ze-studia.
Ngokuvamile, kunezindawo ezintathu zezibalo okudingeka ufunde ngazo:
- Umugqa we-algebra– Ukuze ufunde nge-matrices, ama-eigenvalues kanye nama-veector. Lokhu kusetshenziswa yonke indawo ezindaweni ezinjengezindawo eziphambili zezakhi (i-PCA), i-tensorflow, ngiqonde, ngisho ne-dataframe iyi-matrix!
- Isibaku – Ukuze ufunde ngokuhlukahluka, yilokho ama-algorithms afana ne-gradient feast and backpropagation umsebenzi ngaphansi kwe-hood. Lokhu kusetshenziswe ngokoqobo ngaphakathi kwayo yonke imishini yokufunda i-algorithm yokuziqeqesha nokufunda.
- Izibalo zokubonisa ukuma kwendaba– Ukuqonda amathuba, ukusatshalaliswa, izibalo zaseBayesian, umkhawulo omkhulu we-theorem nokulinganiselwa okuphezulu kokulingana. Izibalo zibaluleke kakhulu kokuphuma kwabathathu, futhi ngangigxila kakhulu ukunaka kwami lapha.
Izinsizakusebenza:
Ngine-athikili egcwele echaza kabanzi ekujuleni kwezihloko zezibalo ozidingayo kanye nokuwohloka okuphelele.
Ungayifunda kanjani izibalo ezidingekayo zokufunda komshini
Python
I-Python yiLingua Franca uma ifika ekufundeni komshini; Khohlwa ngo-R, funda python. (Ngiyaxolisa kubathandi bami bakhona laphaya!)
Ingqikimba ejwayelekile engiyibonile phakathi kwamakhasimende ami okufundisa nabaqalayo engikhulumile nabo ukuthi bazama ukuthola izifundo “ezingcono kakhulu” ukuze bafunde i-Python.
Ngizophinda futhi leli phuzu futhi, “okungcono kakhulu” akulula, ngakho-ke yeka ukuyifuna; kumane kuyindlela yokuhlehlisa. Noma yisiphi isingeniso esidumile se-Python Course sizosebenza, njengoba zizokufundisa izinto ezifanayo.
Noma kunjalo, izinto eziphambili ofuna ukuzifunda yilezi:
- Izakhiwo zedatha yendabuko (Dicts, Tuples, Uhlu)
- Ngoba ngenkathi izihibe
- Uma kungenjalo izitatimende ezinemibandela
- Imisebenzi kanye namakilasi
- Imitapo yolwazi evamile
- Amaphethini wokuqamba
Ufuna futhi ukufunda amaphakheji wokufunda womshini odumile, anjengokuthi:
- Inzotha– Ikhompyutha yenombolo yenombolo yokuhlela.
- Amaphingi ekhanda– Ukukhohlisa kwedatha nokuhlaziywa.
- Matplotlib – Ukubonwa kwedatha kanye nokuhlela.
- Scikit-Funda – Ukuqalisa ukusebenzisa ama-algorithms ayisisekelo.
- Iselele – Iphakheji yesayensi yesayensi ejwayelekile.
Izinsizakusebenza:
I-SQL
Njengonjiniyela wokufunda umshini, uzobe uchitha isikhathi esifanelekile esisebenza ku-SQL lapho uzama ukudala ama-datassets noma wenze ubunjiniyela obuthile.
Cishe ngisebenza ku-SQL ngabo-30- 40% wesikhathi sami njengonjiniyela wokufunda komshini. Kuningi, ngakho-ke nakanjani udinga ukuhlakanipha kwawo, ngaphezu kokucabanga kwakho.
Izinto okufanele zifunde yilezi:
- Khetha * kusuka , Nga-
- Shintsha, faka, dala , Buyekeza , Cisha
- Iqembu ngo, uku-oda ngo
- Lapho, futhi, noma phakathi, ukuba khona
- I-AVG, ukubala, imiz, Max, Isamba
- Joyina Ngokugcwele, Ukujoyina Kwesobunxele, Joyina Ukujoyina, Joyina Ngangaphakathi, Union
- Icala , Iff
- I-Dateadd, DIADIFF, DESTART
- Ukwahlukaniswa, kufanelekile, umugqa ())
Izinsizakusebenza:
Kunezinsiza eziningi zamahhala ze-SQL, ngakho-ke angikuncomi ukuthi uzihluphe ukuchitha imali enkambweni, ngaphandle kokuthi ufuna ngempela. Ungase futhi usebenzise i-chatgpt futhi!
Ukufundwa Komshini
Kumangazo wonke umuntu, kudingeka sifunde ukufunda umshini ukuba ngunjiniyela wokufunda umshini!
Lokhu kuyingxenye ejabulisayo kakhulu ye-RoadMap nokuthi abantu abaningi bangena yini kulo mkhakha. Ngiyakuthola, ngoba bekuyisizathu lapho nginqume ukusebenza ekufundeni komshini!
Ngizobe ngilele uma ngithe ukufunda la ma-algorithms kwakumnandi ngaso sonke isikhathi. Kudinga kancane umzamo wengqondo nesikhathi sokuyiqonda ngokuphelele yonke imiqondo, kepha ekugcineni, izinto zizochofoza, futhi kuzokufanelekela.
Ama-algorithms asemqoka nemiqondo oyidingayo yile:
- Umugqa oqondile, wokubuyiselwa kwemithi kanye ne-polynomial.
- Amamodeli ajwayelekile aqondile namamodeli wezengezo ajwayelekile.
- Izihlahla ezinqunyiwe, amahlathi angahleliwe nezihlahla ezikhuliswe ngeziqu.
- Imishini yokusekela ye-vector.
- K – indlela kanye ne-K-eseduze nomakhelwane.
- Isici sobunjiniyela, ikakhulukazi indlela yokubhekana nezici zesigaba.
- Amamethrikhi wokuhlola ezinhlotsheni ezahlukahlukene zezinkinga.
- Ukujwayelekile, i-Bias vs Variance Trapoff kanye nokuqinisekiswa kwesiphambano.
- I-gradient decent and backpropagation.
Izinsizakusebenza:
- Umshini wokufunda ngokufunda ngo-Andrew Ng – Le yinkambo yokuqala ye-ML engiyithathile, futhi ngicabanga ukuthi mhlawumbe yiyona engcono kakhulu laphaya. U-Andrew uthembekile ngokoqobo uthisha omuhle kakhulu, le nkambo yiwo wonke umuntu okufanele athathe umbono wami.
- Incwadi ye-ML eyikhulu – Kufushane nokuqonda okusebenzayo ekwakheni amamodeli we-ML kanye ne-theory ebalulekile ngemuva kwabo. Ukufundwa kobusuku obuthandekayo.
- Izandla ze-ML nge-Skikit-Funda, Keras, ne-Tessorlow– Ukube bekufanele nginikeze incwadi eyodwa kuphela ukuze ngifunde ukufundwa komshini, lokhu kungaba yiyo! Le ncwadi yimbuzi futhi isembozo ngokoqobo zonke isihloko ongasidinga njengonjiniyela wokufunda / ophakathi komshini webanga eliphakathi.
Ukufunda okujulile
Ukuba qotho, ama-algorithms omshini abayisisekelo azomboza iningi lamamodeli ozowakha emsebenzini wakho.
Ngisasebenzisa amamodeli ajwayelekile wokuhlehlisa isikhathi esiningi!
Ukufunda okujulile kunenzuzo ezimweni ezinjengokwenziwa kolimi lwemvelo kanye nombono wekhompyutha, kepha ukusetshenziswa kwayo emsebenzini wami wansuku zonke kungaphezu kwalezi zindawo.
Kodwa-ke, thatha lokhu ngengcosana kasawoti, unikezwe ukuthi ngisebenza ngokukhethekile ezinkingeni zokubikezela kwezibikezeli zesikhathi kanye nezinkinga zokwandisa, ezikhohlisayo ngokufunda ngokujulile ukuze zenze kahle.
Ngako konke okushiwoyo, ukufundwa okujulile kuyindawo onjiniyela bokufunda umshini kufanele bazi ngandlela thile njengoba kuyingxenye ebalulekile yensimu.
Izindawo ofuna ukuzifunda yilezi:
- Amanethiwekhi we-Neural –I-algorithm ebeka umshini wokufunda ebalazweni. Ngiyaqiniseka ukuthi abaningi kini bezwile ngaleli algorithm.
- Amanethiwekhi we-Neaural nama-Neaural –Lokhu kusetshenziselwa ukubuka kwekhompyutha nokutholwa kwesithombe. Umehluko osemqoka ukuthi basebenzisa umsebenzi wokuzihlukanisa “Khetha kuqala” imininingwane ngaphambi kokuyidlulisela kwinethiwekhi ye-neural ejwayelekile.
- Amanethiwekhi we-neural aphindaphindayo –Kuphele okuncane manje, kepha kwakuyi-algorithm yasekuqaleni yokufunda ejulile yamamodeli wokulandelana njengochungechunge lwesikhathi nolimi lwemvelo. Owathandwa kakhulu okungenzeka ukuthi uke wezwa ngokulandelana kokulandelana kwemibono.
- Abaguquli –Imodeli yamanje yezwe-ye-art esingemuva kwalo lonke i-ai hype nokukhula. Lokhu kuvela ephepheni elidumile elithi “Ukunakwa konke okudingayo”, ukuthi ngincoma kakhulu ukuthi ufunde!
Izinsizakusebenza:
Ubunjiniyela beSoftware
Njengoba kunikezwe isihloko kuwukufunda komshini “unjiniyela”, udinga ukwazi izindlela ezingcono kakhulu zesoftware njengoba lokhu kubalulekile lapho uthumela amamodeli akho ukukhiqiza.
Lapho ngizama ukuba ngunjiniyela wokufunda umshini, ngabukela phansi ingxenye yezobunjiniyela. Ngingaphinde ngiphikisane manje ukuthi kubaluleke kakhulu kunolwazi lomshini wethiyori olufundayo.
Ithiyori rudely nje; Lapho uthola khona imali yakho ngokusiza inkampani nebhizinisi ukwenza izinqumo ngama-algorithms akho. Kulokho, udinga ukwazi ubunjiniyela besoftware.
Izindawo okudingeka wazi ukuthi:
- Izakhiwo zedatha nama-algorithms– Ngokudlulisa izingxoxo kanye nokusiza ukubhala ikhodi engcono. Funda izisekelo futhi uqiniseke ukuthi uzijwayeza.
- Haka
- Uhlu oluxhunyiwe
- Iminyuzi
- Ukuhlunga
- Ukucinga kanambambili
- Izihlahla
- Ukugadwa
- Amagrafu
- Idizayini yesistimu– Ngokudlulisa izingxoxo kanye nokuqonda ukuthi ungathumela kanjani ama-algorithms wokufunda umshini esikalini. Nakulokhu futhi, funda izisekelo.
- Ukuthundla
- Uhlobo lwemvula
- Ukubopha
- Ama-proxies
- Ukubeka
- Ikhodi yokukhiqiza– Ukubhala ikhodi ehlolwe kahle futhi esebenza kahle ngezinto ezifana nokuthayipha, ukubhala, ukuhlola kanye nokusebenzisa imigomo efana ne-DRY, KISS ne-Yagni. Lokhu mhlawumbe kuyingxenye ebaluleke kakhulu okufanele ufunde, ngoba isebenza kakhulu emsebenzini.
- Ama-APIs –Iningi lesoftware lisebenza lisebenzisa ama-API, futhi amamodeli amaningi okufunda umshini anikezwa njenge-API Endpoints. Ukuqonda ukuthi basebenza kanjani nezinhlobo zabo ezihlukile.
Izinsizakusebenza
- Neetmode.io– Isingeniso esikhulu, esakhiwo sedatha esiphakathi nendawo kanye ne-algorithms, kanye nezifundo zokwakhiwa kwesistimu. I-100% itusa le nkundla lapho ufunda okuyisisekelo se-software yobunjiniyela kwanoma ngubani.
- I-LeetCode-Angathi Hackerrrank– Amapulatifomu okuzijwayeza izingxoxo. Ngiyaqiniseka ukuthi abaningi kini bezwile nge “lokugaya leettode”; Awudingi ukwenza lokho ngemisebenzi yonjiniyela yokufunda komshini njengemisebenzi yobunjiniyela yesoftware. Kodwa-ke, kufanele wazi izisekelo. Ngincoma ukusebenza nge-NeetCode 150.
- Ubunjiniyela besoftware ososayensi bedatha– Njengoba isho ku-tin, incwadi eyenzelwe ngqo ososayensi bedatha ukuze bafunde ubunjiniyela besoftware. Enye indlela enhle yokufunda wonke amakhono obunjiniyela besoftware uma ungazithandi izifundo.
Imvelangwane
Imodeli kwi-Jupyter Notebook ayinalo iqiniso lebhizinisi.
Kungcono kakhulu ukuba nokuthile ekukhiqizeni, okwenza izinqumo ezingemuva ezizuzisa ibhizinisi, esikhundleni senethiwekhi ye-Flashy neural kubhukwana ebenza lutho kodwanokunemba okungafanele.
Ngakho-ke, uma empeleni ufuna ukuba ngunjiniyela wokufunda womsindo ozwakalayo, udinga ukukwazi ukufaka amamodeli akho ukuze uzuze inkampani ngombono wezezimali.
Ukuze wenze lokhu, udinga ukufunda okulandelayo:
- Amafu– Funda ubuchwepheshe befu njenge-AWS, GCP noma ama-Azure. Onke amamodeli wokufunda umshini engisebenzele kuwo athunyelwa efwini, futhi lokhu kuzokwanda kuphela esikhathini esizayo. Ama-AWS athandwa kakhulu, ngakho-ke ngincoma ukuthi yilowo ofundayo.
- Isithongo– Funda i-Docker ne-Kubernetes; Lokhu kuyadingeka ukuze kusebenze amamodeli akho efwini.
- Ukulawulwa kwenguqulo– Funda git ne-github, ayikho indlela ezungeze yona. Le yindlela yonke yesoftware eyakhiwe ngayo.
- Igobolondo / terminal– Uzobe usebenza ku-terminal yakho kakhulu, ngakho-ke ukwazi i-bash / zsh eyisisekelo kubalulekile.
Izinsizakusebenza:
- Ama-mlops asebenzayo– Lokhu mhlawumbe yincwadi kuphela okudingeka uyiqonde ukuthi ungayisebenzisa kanjani imodeli yakho yokufunda umshini nazo zonke izihloko ezihambisana nazo. Ngiyisebenzisa kakhulu njengombhalo wereferensi, kepha ifundisa cishe konke okudingeka ukwazi.
- Ukuqamba Izinhlelo Zokufunda Zomshini– Enye incwadi enhle kanye nezinsizakusebenza ukwehluka kwemithombo yakho yolwazi. Lokhu yi-Chip Huyen, okungenzeka ukuthi uchwepheshe oholayo ezinhlelweni zokukhiqiza ze-AI / ML.
- I-Freecodecamp– Izinhlobo ezahlukahlukene zezinsizakusebenza ezimboza ngokoqobo zonke ubunjiniyela besoftware kanye nesihloko se-mlops.
Ukutadisha konke lapha kuzokunikeza lonke ulwazi oludingekayo ukuze ube ngunjiniyela wokufunda umshini; Kodwa-ke, lokho akwanele ngokwayo ukukuyisa emsebenzini.
Udinga ukukhombisa amakhono akho ngokwakha iphothifoliyo eqinile enamaphrojekthi afanele.
Uma ufuna ukwazi kahle ukuthi ungakwenza kanjani lokho, bese ubheka le ndatshana lapho ngichaza khona kahle ukuthi ungakwenza kanjani lokho. Ngizokubona lapho!
Misa ukwakha amaphrojekthi we-ML angenamsebenzi – Okusebenzayo
Enye into!
Nginikela ngezingcingo ezi-1: 1 zokufundisa lapho singaxoxa khona nganoma yini oyidingayo – noma ngabe yimiklamo, izeluleko zomsebenzi, noma nje ukuthola isinyathelo sakho esilandelayo. Ngilapha ukukusiza uqhubekele phambili!
1: 1 Ukulungiswa kwekholi nge-EGOR Howell
Ukuholwa Kwe-Career, Iseluleko Semisebenzi, Usizo Lephrojekthi, Qalisa kabusha Ukubuyekezwaisihloko.io
