I-AIFES: iprojekthi entsha ezisa i-AI kufutshane neArduino

La ibhodi yophuhliso arduino Ivumela amawaka kunye namawaka eeprojekthi ezahlukeneyo ukuba zenziwe, umda ungokoqobo kwingcamango yomenzi ngamnye, nangona ikwanazo ezinye izithintelo zomzimba, ezifana nememori, amandla okusebenza, njl. Nangona kunjalo, kukho iimveliso ezininzi kunye neeprojekthi zokwandisa amandla abo ngakumbi nangakumbi, njengoko kunjalo kwimeko ye ukuqaliswa okutsha kwe-AIFES.

Ngoku, enkosi kule projekthi eyenziwe ngu Fraunhofer IMS for Arduino, le bhodi yemithombo evulekileyo iya kubonisa a I-Artificial Intelligence (AI) isakhelo esicwangciswe kwi-C, usebenzisa iGNU GCC esemgangathweni amathala omqokeleli. Abasebenzisi ngoku baya kuba nakho ukongeza i-AIFES kwiprojekthi yabo yeArduino kwaye bayidibanise kumphathi wethala leencwadi ukusuka kwi-IDE ukuyisebenzisa kuphuhliso lwakho, ukubonelela ngethuba lokusebenzisa ii-algorithms zokufunda koomatshini nakwimicrocontroller ezincinci ezifana nebhodi. Arduino UNO I-8-bit.

Oku kuya kuvumela abaphuhlisi ukuba benze ubuninzi bezixhobo ze-IoT (i-Intanethi yezinto) ezizimeleyo ngakumbi efini kwaye ezinokuba krelekrele ngakumbi, kunye nentlonipho enkulu kubumfihlo bakho, kuba imisebenzi inokuphunyezwa ngaphandle kweintanethi kwibhodi yeArduino ngaphandle kwesidingo. ukuxhomekeka kwiinkonzo ezikude. Ukongeza, iprojekthi ye-AIFES isungulwe phantsi Ilayisensi ye-GNU GPLv3, ngoko ikhululekile ngokupheleleyo, nangona ivumela ilayisenisi ehlawulwayo kwiiprojekthi zorhwebo.

I-AIFES ifana kakhulu kwaye iyahambelana Izicwangciso zePython ML njengokuba kunjalo ngeTensorFlow, Keras okanye iPyTorch, kodwa ukusebenza kwayo kuncitshisiwe. Nangona kunjalo, kule nguqulo ekhutshiweyo i-FNN (i-Feedforward Neural Networks) sele ixhaswe, ngaphezu koko ivumela kwakhona ukusebenza kwemisebenzi edibeneyo efana ne-ReLu, i-Sigmoid, okanye i-Softmax. Kwelinye icala, abaphuhlisi basebenzela ukuzisa kwixesha elizayo kwakhona ukuphunyezwa kweConvNet (Convolutional Neural Networks), enokuthi ingathathi ixesha elide ukufika.

Ezinye nazo zibandakanyiwe algorithms zoqeqesho eziqhelekileyo, ezifana SGD (Gradient Descent Optimizer) kunye Adam Optimizer, phakathi kwabanye. Ndiyathetha, kwi-8-bit MCU, ayimbi kwaphela ...