{"id":41221,"date":"2020-02-06T20:42:46","date_gmt":"2020-02-06T17:42:46","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/python-gateway-v-intersystems-iris"},"modified":"2020-02-06T20:42:46","modified_gmt":"2020-02-06T17:42:46","slug":"python-gateway-v-intersystems-iris","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/python-gateway-v-intersystems-iris","title":{"rendered":"Python Gateway v InterSystems IRIS","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Tento \u010dl\u00e1nek se v\u011bnuje Python Gateway \u2014 komunitn\u00edmu projektu s otev\u0159en\u00fdm zdrojov\u00fdm k\u00f3dem pro datovou platformu InterSystems IRIS. Tento projekt umo\u017e\u0148uje orchestraci jak\u00fdchkoli algoritm\u016f strojov\u00e9ho u\u010den\u00ed vytvo\u0159en\u00fdch v jazyce Python (hlavn\u00edm prost\u0159ed\u00ed pro mnoh\u00e9 datov\u00e9 v\u011bdce), vyu\u017e\u00edvat \u0159adu hotov\u00fdch knihoven pro rychl\u00e9 vytv\u00e1\u0159en\u00ed adaptivn\u00edch, robotizovan\u00fdch analytick\u00fdch AI\/ML \u0159e\u0161en\u00ed na platform\u011b InterSystems IRIS. V tomto \u010dl\u00e1nku uk\u00e1\u017eu, jak m\u016f\u017ee InterSystems IRIS orchestraci proces\u016f v Pythonu efektivn\u011b prov\u00e1d\u011bt oboustrann\u00fd p\u0159enos dat a vytv\u00e1\u0159et inteligentn\u00ed obchodn\u00ed procesy.<\/p>\n

<\/a><\/noindex><\/p>\n

Pl\u00e1n<\/h1>\n

<\/p>\n

    \n
  1. \u00davod.<\/li>\n
  2. N\u00e1stroje.<\/li>\n
  3. Instalace.<\/li>\n
  4. API.<\/li>\n
  5. Interoperabilita.<\/li>\n
  6. Jupyter Notebook.<\/li>\n
  7. Z\u00e1v\u011bry.<\/li>\n
  8. Odkazy.<\/li>\n
  9. MLToolkit.<\/li>\n<\/ol>\n

    <\/p>\n

    \u00davod<\/h1>\n

    <\/p>\n

    Python je vysoko\u00farov\u0148ov\u00fd programovac\u00ed jazyk obecn\u00e9ho ur\u010den\u00ed, kter\u00fd se zam\u011b\u0159uje na zvy\u0161ov\u00e1n\u00ed produktivity v\u00fdvoj\u00e1\u0159\u016f a \u010ditelnost k\u00f3du. V t\u00e9to s\u00e9rii \u010dl\u00e1nk\u016f se budu zab\u00fdvat mo\u017enostmi vyu\u017eit\u00ed jazyka Python na platform\u011b InterSystems IRIS, p\u0159i\u010dem\u017e hlavn\u00edm zam\u011b\u0159en\u00edm tohoto \u010dl\u00e1nku je uplatn\u011bn\u00ed Pythonu jako jazyka pro vytv\u00e1\u0159en\u00ed a aplikaci model\u016f strojov\u00e9ho u\u010den\u00ed.<\/p>\n

    <\/p>\n

    Strojov\u00e9 u\u010den\u00ed (ML) je t\u0159\u00edda metod um\u011bl\u00e9 inteligence, kter\u00e9 se vyzna\u010duj\u00ed t\u00edm, \u017ee p\u0159\u00edmo ne\u0159e\u0161\u00ed probl\u00e9m, ale u\u010d\u00ed se p\u0159i \u0159e\u0161en\u00ed mnoha podobn\u00fdch \u00faloh. <\/p>\n

    <\/p>\n

    Algoritmy a modely strojov\u00e9ho u\u010den\u00ed se st\u00e1vaj\u00ed st\u00e1le b\u011b\u017en\u011bj\u0161\u00edmi. D\u016fvod\u016f je mnoho, ale v\u0161echno se to vrac\u00ed k dostupnosti, jednoduchosti a dosa\u017een\u00ed praktick\u00fdch v\u00fdsledk\u016f. Je klasifikace nebo dokonce neuronov\u00e9 modelov\u00e1n\u00ed novou technologi\u00ed? <\/p>\n

    <\/p>\n

    Samoz\u0159ejm\u011b, \u017ee ne, ale v sou\u010dasnosti nen\u00ed pot\u0159eba ps\u00e1t stovky tis\u00edc \u0159\u00e1dk\u016f k\u00f3du, abychom spustili jeden model, a n\u00e1klady na vytv\u00e1\u0159en\u00ed a aplikaci model\u016f neust\u00e1le klesaj\u00ed.<\/p>\n

    <\/p>\n

    N\u00e1stroje se vyv\u00edjej\u00ed \u2014 zat\u00edmco nem\u00e1me zcela GUI-orientovan\u00e9 AI\/ML n\u00e1stroje, pokrok, kter\u00fd jsme zaznamenali u mnoha dal\u0161\u00edch t\u0159\u00edd informa\u010dn\u00edch syst\u00e9m\u016f, nap\u0159\u00edklad BI (od psan\u00ed k\u00f3du po pou\u017e\u00edv\u00e1n\u00ed r\u00e1mc\u016f a GUI-orientovan\u00fdch konfigurovateln\u00fdch \u0159e\u0161en\u00ed), se objevuje i u n\u00e1stroj\u016f pro vytv\u00e1\u0159en\u00ed AI\/ML. Ji\u017e jsme p\u0159ekonali f\u00e1zi psan\u00ed k\u00f3du a dnes pou\u017e\u00edv\u00e1me r\u00e1mce k budov\u00e1n\u00ed a tr\u00e9nov\u00e1n\u00ed model\u016f.<\/p>\n

    <\/p>\n

    Dal\u0161\u00ed vylep\u0161en\u00ed, jako je mo\u017enost distribuce p\u0159edtr\u00e9novan\u00e9ho modelu, kdy koncov\u00fd u\u017eivatel mus\u00ed pouze dokon\u010dit tr\u00e9nink modelu na sv\u00fdch specifick\u00fdch datech, tak\u00e9 usnad\u0148uj\u00ed za\u010d\u00e1tek uplatn\u011bn\u00ed strojov\u00e9ho u\u010den\u00ed. Tyto pokroky v\u00fdrazn\u011b zjednodu\u0161uj\u00ed studium strojov\u00e9ho u\u010den\u00ed jak pro odborn\u00edky, tak pro spole\u010dnosti jako celek.<\/p>\n

    <\/p>\n

    Na druh\u00e9 stran\u011b sb\u00edr\u00e1me st\u00e1le v\u00edce dat. D\u00edky jednotn\u00e9 datov\u00e9 platform\u011b, jako je InterSystems IRIS, mohou b\u00fdt v\u0161echny tyto informace okam\u017eit\u011b p\u0159ipraveny a pou\u017eity jako v\u00fdchoz\u00ed data pro modely strojov\u00e9ho u\u010den\u00ed.<\/p>\n

    <\/p>\n

    S p\u0159echodem do cloudu je spu\u0161t\u011bn\u00ed AI\/ML projekt\u016f snaz\u0161\u00ed ne\u017e kdykoli p\u0159edt\u00edm. M\u016f\u017eeme spot\u0159ebov\u00e1vat pouze ty zdroje, kter\u00e9 pot\u0159ebujeme. Nav\u00edc d\u00edky paralelizaci, kterou nab\u00edzej\u00ed cloudov\u00e9 platformy, m\u016f\u017eeme u\u0161et\u0159it \u010das.<\/p>\n

    <\/p>\n

    Ale co v\u00fdsledky? Tady se situace komplikuje. Existuje mnoho n\u00e1stroj\u016f pro tvorbu model\u016f, o kter\u00fdch budu hovo\u0159it d\u00e1le. Vytvo\u0159it dobr\u00fd model nen\u00ed snadn\u00e9, ale co d\u00e1l? Z\u00edsk\u00e1n\u00ed zisku z pou\u017eit\u00ed modelu firmou je tak\u00e9 netrivi\u00e1ln\u00ed \u00fakol. J\u00e1dr\u016fm probl\u00e9mu je rozd\u011blen\u00ed analytick\u00fdch a transak\u010dn\u00edch z\u00e1t\u011b\u017e\u00ed a datov\u00fdch model\u016f. Kdy\u017e tr\u00e9nujeme model, obvykle to d\u011bl\u00e1me na historick\u00fdch datech. Ale m\u00edsto, kde m\u00e1 postaven\u00fd model fungovat, je v transak\u010dn\u00edm zpracov\u00e1n\u00ed dat. Jak\u00fd m\u00e1 smysl nejlep\u0161\u00ed model pro detekci podvodn\u00fdch transakc\u00ed, pokud ho spou\u0161t\u00edme jednou za den? Podvodn\u00edci u\u017e d\u00e1vno utekli s pen\u011bzi. Mus\u00edme tr\u00e9novat model na historick\u00fdch datech, ale tak\u00e9 ho mus\u00edme aplikovat v re\u00e1ln\u00e9m \u010dase na nov\u00fdch p\u0159ich\u00e1zej\u00edc\u00edch datech, aby na\u0161e podnikov\u00e9 procesy mohly jednat v souladu s predikcemi proveden\u00fdmi modelem.<\/p>\n

    <\/p>\n

    ML Toolkit je soubor n\u00e1stroj\u016f, jeho\u017e c\u00edlem je pr\u00e1v\u011b toto: spojen\u00ed model\u016f a transak\u010dn\u00edho prost\u0159ed\u00ed, aby bylo mo\u017en\u00e9 pou\u017eit\u00ed vytvo\u0159en\u00fdch model\u016f p\u0159\u00edmo ve va\u0161ich obchodn\u00edch procesech. Python Gateway je sou\u010d\u00e1st\u00ed ML Toolkit a zaji\u0161\u0165uje integraci s jazykem Python (podobn\u011b jako R Gateway, kter\u00e1 je sou\u010d\u00e1st\u00ed ML Toolkit, zaji\u0161\u0165uje integraci s jazykem R).<\/p>\n

    <\/p>\n

    N\u00e1stroje<\/h1>\n

    <\/p>\n

    Ne\u017e budeme pokra\u010dovat, r\u00e1d bych popsal n\u011bkolik n\u00e1stroj\u016f a knihoven pro Python, kter\u00e9 budeme pou\u017e\u00edvat pozd\u011bji.<\/p>\n

    <\/p>\n

    Technologie<\/h2>\n

    <\/p>\n