Ukusebenzisa umtapo wezincwadi wekhompuyutha onamandla wesayensi, i-Scipy, kuhlanganisa ukuqonda intaphane yemisebenzi engaxazulula izinkinga eziningi. Umsebenzi owodwa onjalo ukuqaliswa kwe-Scpy ekubaleni i-Kullback-Leibler Divergence. Njengokubuka konke, ukwehluka kwe-Kullback-Leibler isilinganiso sokuthi ukusabalalisa kwamathuba okukodwa kwehluka kanjani kwesekhondi, okulindelwe ukusabalalisa kwamathuba.
I-Kullback-Leibler Divergence
I-Kullback-Leibler Divergence (KLD) isetshenziswa kakhulu ezimeni ezihlanganisa ukufunda komshini kanye nethiyori yolwazi, njengendlela yokulinganisa umehluko phakathi kokusabalalisa kwamathuba eqiniso nokubikezelwe. Ikakhulukazi, ivamise ukusetshenziswa ezinkingeni zokuthuthukisa lapho inhloso kuwukunciphisa umehluko phakathi kokusatshalaliswa okubikezelwe nokwangempela.
Ku-Python, ikakhulukazi ngaphakathi komtapo wezincwadi we-Scpy, i-Kullback-Leibler Divergence isetshenziswa kukho kokubili ukusabalalisa okuqhubekayo nokuhlukile.
Le ndlela yenza lula kakhulu inqubo yokubala kokuhlukana, ukunciphisa isidingo sokusebenzisa mathupha ama-algorithms ezibalo, noma ukubhekana nobunzima bokuhlanganisa izinombolo.
I-Scipy Kullback-Leibler Divergence
Ukuze ufanekisele ukwehlukana kwe-Scipy Kullback-Leibler, sizokhiqiza ukusabalalisa okubili kwamathuba futhi sibale umehluko phakathi kwakho.
import numpy as np from scipy.special import kl_div # Generate distributions p = np.array([0.1, 0.2, 0.3, 0.4]) q = np.array([0.3, 0.2, 0.2, 0.3]) # Calculate KL Divergence kl_divergence = kl_div(p,q).sum() print(kl_divergence)
Lesi sibonelo singenisa kuqala imitapo yolwazi edingekayo. Sichaza amalungu afanayo amabili, ngalinye limelela ukusabalalisa kwamathuba ahlukene (p kanye no-q). I-Kullback-Leibler Divergence ibe isibalwa kusetshenziswa umsebenzi othi `kl_div` ukusuka kumojuli `scipy.special`, ebuyisela amalungu afanayo anobude obufanayo. Isamba salolu hlu singukwehluka kwe-KL okuphelele.
Ukutolika Imiphumela
Ukuze uqonde imiphumela evela ekusetshenzisweni kwe-Scipy kwe-KLD, kubalulekile ukuqaphela ukuthi ukwehlukana akusona ncamashi isilinganiso “sebanga” njengoba singalingani. Lokhu kusho ukuthi ukwehlukana kwe-KL kuka-P ku-Q akufani nokuhlukana kwe-KL kuka-Q kusukela ku-P.
Ngakho, uma ukwehlukana kwe-KL okubaliwe kukuncane, iphakamisa ukuthi ukusabalalisa u-P no-Q kuyafana okunye. Ngokuphambene, ukwehlukana okuphezulu kwe-KL kusho ukuthi ukusatshalaliswa kwehluka kakhulu.
Ezinkingeni zokufunda nokuthuthukisa umshini, inhloso ngokuvamile ukushuna amapharamitha emodeli ukuze kuncishiswe ukwehlukana kwe-KL, okuholela kumodeli engabikezela ukusatshalaliswa eduze nokusatshalaliswa kwangempela.
Ukuhlola Okwengeziwe
I-Scipy inikeza inqwaba yamakhono nezixazululo zezinkinga zezibalo eziyinkimbinkimbi ezingaphezu kwesibalo se-Kullback-Leibler Divergence. Ngaphandle kwesibalo sokuhluka, kuneminye imisebenzi eminingi yezibalo nezibalo njengokuhlanganiswa, ukuhumusha, ukwenza kahle, ukucutshungulwa kwesithombe, i-algebra yomugqa nokunye. Ukuwasebenzisa kungenza kube lula kakhulu inqubo yokwakheka kwe-algorithm nokuhlaziywa kwedatha.
Leli ithuluzi elibalulekile kunoma ubani abathintekayo kukhompyutha yesayensi, isayensi yedatha, noma ukufunda ngomshini, futhi kusiza ukuqeda ubunkimbinkimbi obukhona ekusetshenzisweni mathupha, kukuvumela ukuthi uthuthukise futhi ulungiselele izixazululo zakho ngempumelelo kakhudlwana.
