Sawubona, Habr! Ngethula ekunakeni kwakho ukuhunyushwa kwalesi sihloko ngu-Kristencarter7519.
Nakuba kubonakala sengathi sisekude nezinsuku ezimbalwa kusukela ngo-2020, lezi zinsuku zibalulekile futhi emkhakheni wokuthuthukiswa kwesofthiwe. Lapha kulesi sihloko, sizobona ukuthi unyaka ozayo ka-2020 uzoshintsha kanjani impilo yabathuthukisi be-software.
Ikusasa lokuthuthukiswa kwesofthiwe selifikile!
Ukuthuthukiswa kwesofthiwe yendabuko ukuthuthukiswa kwesofthiwe ngokubhala ikhodi ngokulandela imithetho ethile engashintshi. Kodwa ukuthuthukiswa kwesofthiwe yesimanje kuye kwafakazela ukushintsha kwepharadigm nokuthuthuka kobuhlakani bokwenziwa, ukufunda ngomshini nokufunda okujulile. Ngokuhlanganisa lobu buchwepheshe obuthathu, abathuthukisi bazokwazi ukudala izixazululo zesofthiwe ezifunda emiyalweni futhi bengeze izici ezengeziwe namaphethini kudatha edingekayo ukuze kukhiqizwe umphumela oyifunayo.
Ake sizame ngekhodi ethile
Ngokuhamba kwesikhathi, izinhlelo zokuthuthukisa isofthiwe yenethiwekhi ye-neural ziye zaba inkimbinkimbi ngokwemibandela yokuhlanganisa kanye namazinga okusebenza nezixhumi ezibonakalayo. Abathuthukisi, isibonelo, bangakha inethiwekhi ye-neural elula kakhulu ngePython 3.6. Nasi isibonelo sohlelo olwenza ukuhlelwa kanambambili ngo-1 noma u-0.
Kunjalo, singaqala ngokwakha isigaba senethiwekhi ye-neural:
ngenisa i-NumPy njenge-NP
X=np.array([[0,1,1,0],[0,1,1,1],[1,0,0,1]])
y=np.array([[0],[1],[1]])Ukusetshenziswa komsebenzi we-sigmoid:
def sigmoid ():
return 1/(1 + np.exp(-x))
def derivatives_sigmoid ():
return x * (1-x)Ukuqeqesha imodeli enesisindo sokuqala nokuchema:
epoch=10000
lr=0.1
inputlayer_neurons = X.shape[1]
hiddenlayer_neurons = 3
output_neurons = 1
wh=np.random.uniform(size=(inputlayer_neurons,hiddenlayer_neurons))
bh=np.random.uniform(size=(1,hiddenlayer_neurons))
wout=np.random.uniform(size=(hiddenlayer_neurons,output_neurons))
bout=np.random.uniform(size=(1,output_neurons))Kwabaqalayo, uma udinga usizo mayelana namanethiwekhi e-neural, ungasesha ku-inthanethi amawebhusayithi ezinkampani eziphezulu zokuthuthukisa isofthiwe noma ungaqasha onjiniyela be-AI/ML ukuthi basebenze kuphrojekthi yakho.
Ukuguqulwa kwekhodi kusetshenziswa i-neuron yesendlalelo esiphumayo
hidden_layer_input1=np.dot(X,wh)
hidden_layer_input=hidden_layer_input1 + bh
hiddenlayer_activations = sigmoid(hidden_layer_input)
output_layer_input1=np.dot(hiddenlayer_activations,wout)
output_layer_input= output_layer_input1+ bout
output = sigmoid(output_layer_input)
Iphutha lokubala lesendlalelo sekhodi efihliwe
E = y-output
slope_output_layer = derivatives_sigmoid(output)
slope_hidden_layer = derivatives_sigmoid(hiddenlayer_activations)
d_output = E * slope_output_layer
Error_at_hidden_layer = d_output.dot(wout.T)
d_hiddenlayer = Error_at_hidden_layer * slope_hidden_layer
wout += hiddenlayer_activations.T.dot(d_output) *lr
bout += np.sum(d_output, axis=0,keepdims=True) *lr
wh += X.T.dot(d_hiddenlayer) *lr
bh += np.sum(d_hiddenlayer, axis=0,keepdims=True) *lrPhuma
print (output)
[[0.03391414]
[0.97065091]
[0.9895072 ]]Kuhlale kufanelekile ukuhlala unolwazi lwakamuva ngezilimi zokuhlela zakamuva nezindlela zokubhala amakhodi, futhi abahleli bohlelo kufanele baqaphele amathuluzi amaningi amasha asiza ukwenza izinhlelo zabo zokusebenza zihambisane nabasebenzisi abasha.
Ngo-2020, abathuthukisi be-software kufanele bacabangele ukufaka lawa mathuluzi angu-5 okuthuthukisa isofthiwe emikhiqizweni yabo, kungakhathaliseki ukuthi basebenzisa luphi ulimi lohlelo:
1. Ukucubungula Ulimi Lwemvelo (NLP)
Nge-chatbot eyenza lula isevisi yamakhasimende, i-NLP ithola ukunakwa kwabahleli bezinhlelo abasebenza ekuthuthukisweni kwesofthiwe yesimanje. Basebenzisa amathuluzi e-NLTK afana ne-Python NLTK ukuze bafake ngokushesha i-NLP kuma-chatbots, abasizi bedijithali, nemikhiqizo yedijithali. Maphakathi no-2020 noma maduze nje, uzobona i-NLP iba ββebaluleke kakhulu kukho konke kusukela kumabhizinisi okuthengisa kuya ezimotweni ezizimele kanye namadivayisi wekhaya nehhovisi.
Ukuqhubekela phambili ngamathuluzi angcono okuthuthukisa isofthiwe nobuchwepheshe, ungalindela ukuthi abathuthukisi besofthiwe basebenzise i-NLP ngezindlela ezihlukahlukene, kusukela ekuxhumaneni komsebenzisi okusekelwe ezwini kuya ekuzulazuleni kwemenyu okulula kakhulu, ukuhlaziya imizwa, ukuhlonza umongo, imizwa, nokufinyeleleka kwedatha. Konke lokhu kuzotholakala kubasebenzisi abaningi, futhi izinkampani zizokwazi ukuzuza ukukhula kokukhiqiza okufika ku-$430 billion ngo-2020 (ngokwe-IDC, ecashunwe yi-Deloitte).
2. I-GraphQL ithatha indawo ye-REST Apis
Ngokusho konjiniyela befemu yami, okuyinkampani yokuthuthukisa isofthiwe ye-offshore, i-REST API ilahlekelwa ukubusa kwayo endaweni yonke yohlelo lokusebenza ngenxa yokulayishwa kancane kwedatha okudingeka yenziwe kuma-URL amaningi ngawodwana.
I-GraphQL iyithrendi entsha kanye nenye indlela engcono yezakhiwo ezisekelwe ku-REST ebuyisa yonke idatha efanelekile kumasayithi amaningi kusetshenziswa umbuzo owodwa. Lokhu kuthuthukisa ukusebenzisana kweklayenti-neseva futhi kunciphisa ukubambezeleka, okwenza uhlelo lokusebenza luphendule kakhulu kumsebenzisi.
Ungathuthukisa amakhono akho okuthuthukisa isofthiwe uma usebenzisa i-GraphQL ekuthuthukiseni isofthiwe. Ukwengeza, idinga ikhodi encane kune-REST Api futhi ikuvumela ukuthi wenze imibuzo eyinkimbinkimbi ngemigqa embalwa elula. Ingase futhi ifakwe izici eziningi ze-Backand as a Service (BaaS) ezenza kube lula ukusetshenziswa ngabathuthukisi be-software ngezilimi ezihlukahlukene zokuhlela, okuhlanganisa i-Python, i-Node.js, i-C++ ne-Java.
3. Izinga lekhodi ephansi/ayikho ikhodi (ikhodi ephansi)
Wonke amathuluzi okuthuthukisa isofthiwe yekhodi ephansi ahlinzeka ngezinzuzo eziningi. Kufanele isebenze kahle ngangokunokwenzeka lapho ubhala izinhlelo eziningi kusukela ekuqaleni. Ikhodi ephansi inikeza ikhodi elungiselelwe ngaphambili engashumeka ezinhlelweni ezinkulu. Lokhu kuvumela nabangezona izinhlelo ukuthi bakhe ngokushesha futhi kalula imikhiqizo eyinkimbinkimbi futhi basheshise i-ecosystem yesimanje yokuthuthukiswa.
Ngokombiko we-TechRepublic, amathuluzi e-no-code/low code asesetshenziswa kakade kuma-portal ewebhu, izinhlelo zesofthiwe, izinhlelo zokusebenza zeselula nezinye izindawo. Imakethe yamathuluzi ekhodi ephansi izokhula ifinyelele ku-$15 billion ngo-2020. Lawa mathuluzi aphatha yonke into, okuhlanganisa ukuphatha ingqondo yokuhamba komsebenzi, ukuhlunga idatha, ukungenisa nokuthekelisa. Nazi izinkundla ezingcono kakhulu zamakhodi aphansi ngo-2020:
- I-Microsoft PowerApps
- I-Mendix
- Amasistimu angaphandle
- UMdali we-Zoho
- Salesforce App Cloud
- Isizinda Esisheshayo
- Ibhuthi yasentwasahlobo
4G igagasi
Ukuxhumana kwe-5G kuzoba nomthelela omkhulu ohlelweni lokusebenza lweselula nokuthuthukiswa kwesofthiwe kanye nokuthuthukiswa kwewebhu. Phela, ngobuchwepheshe obufana ne-IoT, yonke into ixhunyiwe. Ngakho-ke, isofthiwe yedivayisi izokwenza ngokugcwele amandla amanethiwekhi angenantambo anesivinini esiphezulu nge-5G.
Engxoxweni yakamuva ne-Digital Trends, uDan Dery, iphini likamongameli we-Motorola womkhiqizo, uthe "eminyakeni ezayo, i-5G izoletha idatha esheshayo, i-bandwidth ephakeme, futhi isheshise isofthiwe yocingo izikhathi ezingu-10 ngokushesha kunobuchwepheshe obukhona obungenawaya."
Ngalokhu kukhanya, izinkampani zesoftware zizosebenza ukuletha i-5G kuzinhlelo zokusebenza zesimanje. Njengamanje, opharetha abangaphezu kuka-20 bamemezele ukuthuthukiswa kwamanethiwekhi abo. Ngakho-ke, abathuthukisi manje bazoqala ukusebenza ekusebenziseni ama-API afanelekile ukuze bazuze i-5G. Ubuchwepheshe buzothuthukisa kakhulu okulandelayo:
- Ukuphepha kohlelo lwenethiwekhi, ikakhulukazi Ukusika Kwenethiwekhi.
- Nikeza ngezindlela ezintsha zokuphatha ama-ID omsebenzisi.
- Ikuvumela ukuthi ungeze ukusebenza okusha ezinhlelweni zokusebenza ezinokulibaziseka okuphansi.
- Izoba nomthelela ekuthuthukisweni kwesistimu ye-AR/VR.
5. Ukuqinisekisa okulula
Ukuqinisekisa kuya ngokuya kuba inqubo esebenzayo yokuvikela idatha ebucayi. Ubuchwepheshe obuyinkimbinkimbi abugcini nje ngokuba sengozini yokugetshengwa kwesoftware, kodwa futhi busekela ubuhlakani bokwenziwa kanye ne-quantum computing. Kodwa imakethe yokuthuthukisa isofthiwe isivele ibona izinhlobo eziningi ezintsha zokuqinisekisa, njengokuhlaziywa kwezwi, i-biometrics nokuqashelwa kobuso.
Kulesi sigaba, abaduni bathola izindlela ezihlukene zokwenza ama-ID wabasebenzisi be-inthanethi namaphasiwedi. Njengoba abasebenzisi beselula sebejwayele ukufinyelela kuma-smartphones abo ngezigxivizo zeminwe noma zobuso, ngaleyo ndlela besebenzisa amathuluzi okuqinisekisa, ngeke badinge amakhono amasha okuqinisekisa njengoba amathuba okuntshontshwa kwe-inthanethi azoba mancane. Nawa amanye amathuluzi okufakazela ubuqiniso bezinto eziningi ngokubethela kwe-SSL.
- Ama-Soft Tokens aguqula ama-smartphones akho abe iziqinisekisi zezinto eziningi ezisebenzayo.
- Izifanekiso ze-EGrid ziyindlela elula yokusebenzisa futhi edumile yokuqinisekisa embonini.
- Ezinye zezinhlelo ezingcono kakhulu zokuqinisekisa zamabhizinisi iRSA SecurID Access, i-OAuth, i-Ping Identity, i-Authx, ne-Aerobase.
Kunezinkampani zama-software e-India nase-US ezenza ucwaningo olunzulu emkhakheni wokuqinisekisa kanye ne-biometrics. Baphinde bakhuthaze i-AI ukuze yenze isofthiwe ephakeme kakhulu yezwi, i-face-id, ukuziphatha kanye nokuqinisekiswa kwe-biometric. Manje usungakwazi ukuvikela iziteshi zedijithali futhi uthuthukise amakhono enkundla.
isiphetho
Kubukeka sengathi impilo yabahleli bezinhlelo izoba yinselelo encane ngo-2020 njengoba ijubane lokuthuthukiswa kwesoftware kungenzeka lisheshe. Amathuluzi atholakalayo azoba lula ukuwasebenzisa. Ekugcineni, le ntuthuko izodala umhlaba onamandla ongena enkathini entsha yedijithali.
Source: www.habr.com