Misali na Tambayoyin Tattaunawa don Matsayin Metropolis
A cikin mahallin kwaikwayon Monte Carlo, matakin Metropolis muhimmin tsari ne a cikin tsarin kididdiga da sauran fannoni. A cikin wannan sashe, mun tattauna musamman hanyar Metropolis-Hastings, wani tsari da ake amfani da shi don yin samfuri daga rarrabawar yiwuwar rikitarwa. Ta hanyar fahimtar matakan da ke cikin wannan tsari, za mu iya yin kwaikwayon da ya fi daidai kuma mai inganci.
Gabatarwa ga Tsarin Tsarin Metropolis
Nicholas Metropolis da abokan aikinsa ne suka gabatar da tsarin Metropolis a shekarar 1953. Ana amfani da wannan hanyar don yin koyi da kuma kwaikwayon yanayin tsarin jiki, musamman waɗanda suka shafi barbashi da yawa kamar iskar gas ko ruwa. Sigar zamani ta wannan tsarin, Metropolis-Hastings, wani tsari ne na gabaɗaya wanda ke ba da damar zana samfura daga rarrabawa mai manufa mara tsari.
Matakai a cikin Tsarin Algorithm na Metropolis
Domin fahimtar yadda tsarin aikin Metropolis yake aiki, yana da mahimmanci a fahimci matakan:
1. Farawa: Fara da zaɓar mafita ta farko bazuwar daga sararin maganin ko rarrabawa ta farko. Misali, muna farawa da yanayin zafi ko matsayin barbashi.
2. Shawarar Sabuwar Mataki: Shawarar sabuwar yanayi (sabon mafita) ta hanyar yin ƙaramin canji ga yanayin da ake ciki. Sau da yawa ana kiran wannan da matakin "shawarar". Wannan canjin yawanci ana ɗauko shi ne daga rarrabawa mai daidaito, kamar rarrabawar Gaussian.
3. Lissafin Rabon Karɓa: Lissafin rabon karɓa, wanda ke ƙayyade ko mun yarda ko mun ƙi wani motsi da aka gabatar. Wannan rabo shine rabon yuwuwar sabuwar yanayin zuwa yanayin da ake ciki. A cikin bayanin lissafi, an bayar da wannan rabo ta hanyar:
\[
A = \min\left(1, \frac{P(\text{new})}{P(\text{current})}\right)
\]
inda \( P \) shine yuwuwar wani yanayi na musamman.
4. Shawara Ta Amfani da Rabon Karɓa: Kwatanta rabon karɓa da ƙimar bazuwar da aka ɗauko daga rarrabawa iri ɗaya tsakanin 0 da 1. Idan rabon karɓa ya fi ƙimar bazuwar girma, karɓi sabon motsi; in ba haka ba, ƙi shi kuma ka ci gaba da kasancewa a cikin yanayin da ake ciki.
5. Maimaitawar: Maimaita matakai 2 zuwa 4 don adadin maimaitawar da ake so ko har sai tsarin ya kai daidaito.
Tambayoyi da Tattaunawa Samfura
Bari mu tattauna wasu misalai na tambayoyi domin mu fahimci matakin Metropolis sosai.
Misali Tambaya ta 1
Tambaya: Kuna da barbashi a cikin girma ɗaya na matsayi \( x \) wanda aikin makamashi mai yuwuwa \( U(x) = x^2 \) ke shafar. Yi amfani da algorithm na Metropolis don kwaikwayon rarraba matsayin barbashi.
Tattaunawa:
1. Farawa: Fara daga matsayi \( x = 0 \).
2. Ba da Shawarar Sabuwar Motsi: Ba da Shawarar sabon matsayi \( x' = x + \Delta x \), tare da \( \Delta x \) da aka ɗauko daga rarrabawar Gaussian tare da matsakaicin sifili.
3. Lissafin Rabon Makamashi: Lissafa rabon makamashi:
\[
\Delta U = U(x') – U(x) = x'^2 – x^2
\]
Don haka, rabon karɓa shine:
\[
A = \min\left(1, e^{-\Delta U}\right)
\]
4. Shawara: Idan \( A \) ya fi lamba tsakanin 0 da 1, karɓa \( x' \); in ba haka ba, tsaya a matsayin \( x \).
5. Maimaitawa: Maimaita wannan tsari a matakai 10,000.
Rarraba matsayi da aka samu zai biyo bayan rarrabawar Gaussian tare da sifili matsakaici da bambancin da ya yi daidai da yuwuwar, wanda a wannan yanayin, ke haifar da rarrabawa wanda aikin makamashi mai yuwuwa ya tsara.
Misali Tambaya ta 2
Tambaya: Yi amfani da tsarin Metropolis don dacewa da ƙa'idar aikin Bayesian. Bari mu ce muna son daidaita gangara mai sauƙi a cikin bayanai ta amfani da layi mai juyawa tare da MCMC.
Tattaunawa:
1. Farawa: Saita sigogin samfurin farko \( \beta = (m, c) \).
2. Shawarar Sabon Mataki: Ba da shawarar sabbin sigogi na rarraba shawarwari na yau da kullun na multivariate. Misali, yi amfani da rarraba Gaussian don masu canji \( m \) da \( c \).
3. Rabon Karɓa: Lissafa rabon karɓa ta hanyar:
\[
A = \min\left(1, \frac{L(m', c'| \text{data})P(m', c')}{L(m, c| \text{data})P(m, c)}\right)
\]
Inda \( L \) shine yuwuwar, kuma \( P \) shine gaban siga.
4. Shawara: Kwatanta rabon da ƙimar bazuwar 0 zuwa 1 don karɓa ko ƙin amincewa da shawarar.
5. Maimaitawa: Gudanar da kwaikwayon tare da isasshen maimaitawa har sai an cimma haɗuwa.
Da wannan hanyar, za mu iya samun rarrabawar baya don sigogin koma-baya, wanda ke ba mu hanyar fahimta da fassara alaƙar da ke cikin bayanai.
Kammalawa
Matakan Metropolis a cikin kwaikwayon Monte Carlo yana ba mu damar yin samfuri daga rarrabawa masu rikitarwa kuma yana aiki a matsayin tushen hanyar Metropolis-Hastings. Ta hanyar amfani da wannan dabarar zuwa fannoni daban-daban, za mu iya cimma ingantaccen samfurin ƙira da fahimtar tsarin dalla-dalla. A cikin aikace-aikace tun daga kimiyyar lissafi da ilmin halitta zuwa kimiyyar kwamfuta da ƙididdiga, wannan algorithm yana ba da mafita masu kyau da tasiri ga matsaloli masu rikitarwa.
