Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Waainishaji wa Naive Bayes

Angalia jinsi Naive Bayes inavyochanganya vipaumbele vya darasa na uwezekano wa vipengele, na jinsi vibadala vyake hushughulikia hesabu, vipengele vya mfumo wa jozi na data endelevu.

Njia za mwongozo zinazohusianaWaainishaji wa Bayes wasio na akili
Swali 1 ya 8

Ni nini dhana ya msingi ya "kutojua" katika darasa la Naive Bayes?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Ni nini dhana ya msingi ya "kutojua" katika darasa la Naive Bayes?
    • Vipengele vinajitegemea kwa masharti kutokana na darasa
    • Madarasa yana uwezekano sawa
    • Vipengele vyote vinaendelea na kawaida husambazwa
    • Seti ya data imesawazishwa kikamilifu
  2. Naive Bayes imejengwa juu ya nadharia gani?
    • Nadharia ya kikomo cha kati
    • Bayes’ theorem
    • Sheria ya idadi kubwa
    • The Cauchy-Schwarz inequality
  3. Kwa nini utekelezaji kawaida hujumlisha uwezekano wa logi badala ya kuzidisha uwezekano mbichi?
    • Ili kufanya mfano usiwe wa mstari
    • Ili kuharakisha mafunzo kwa kupita moja
    • Ili kuzuia mtiririko mdogo wa nambari kutoka kwa kuzidisha nambari nyingi ndogo
    • Ili kutekeleza uhuru wa kipengele
  4. Je, Laplace (nyongeza-moja) inasuluhisha tatizo gani?
    • Inasawazisha ukubwa wa darasa
    • Inarekebisha vipengele vinavyoendelea
    • Huondoa vipengele vinavyohusiana
    • Huzuia kipengele kisichoonekana kulazimisha uwezekano wa sifuri
  5. Ni lahaja gani ya Naive Bayes ambayo ni ya asili zaidi kwa vipengele vya kuhesabu maneno kwenye hati?
    • Multinomial Naive Bayes
    • Gaussian Naive Bayes
    • Bernoulli Naive Bayes
    • Linear Naive Bayes
  6. Why might Naive Bayes be useful as a lightweight baseline?
    • Inahakikisha usahihi bora kwenye mkusanyiko wowote wa data
    • Ni haraka, inatafsirika, na inafanya kazi vizuri na data ndogo
    • Hujifunza kiotomatiki mwingiliano wa vipengele
    • Haihitaji data ya mafunzo
  7. What distribution does Gaussian Naive Bayes assume for each feature within a class?
    • A uniform distribution over every possible value
    • A multinomial distribution over word counts
    • A Gaussian (normal) distribution
    • A distribution that is always exactly zero
  8. Which feature representation fits Bernoulli Naive Bayes most naturally?
    • The number of times each word occurs
    • A continuous pixel intensity modeled by a normal curve
    • A sequence of audio frames
    • Whether each word is present or absent in a document