Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application



Download 1,94 Mb.
Pdf ko'rish
bet77/96
Sana22.06.2022
Hajmi1,94 Mb.
#692449
1   ...   73   74   75   76   77   78   79   80   ...   96
Bog'liq
2021272010247334 5836879612033894610

Binary Classification: “Let ‘Z = X × {0,1}’, let ‘H’ be a set of functions
‘h: X → {0,1}’, and let ‘f (h; (x, y)) = 11
{h(x)̸ =y}
. Here, ‘f (·)’ is simply the ‘0
− 1 loss function’, measuring whether the binary hypothesis ‘h(·)’
misclassified the example (x,y)”.
This setting’s ultimate goal is a selection of a hypothesis 

 H” based on
a finite number of samples with the least amount of potential risk. In
general, we are expecting that the sample size will improve the
approximation of the risk. It is assumed that “learning guidelines that
enable us to choose such hypotheses are consistent”. In formal terms, we
conclude that “rule A” is consistent with rate “εcons(m)” under distribution
D” if for all “m”, where “F* = inf
h


F(h)”, the “rate” ε(m) is required to
be monotone decreasing with “εcons(m) −→ 0)”.
“ES

Dm [F(A(S)) − F

] ≤ εcons(m)”
We can not choose a "D-based" learning rule because "D" is unknown.
Rather, we need a “stronger requirement that the rule is consistent with rate


εcons(m) under all distributions over Z”. The key definition is as follows:
“A learning problem is learnable, if there exist a learning rule A and a
monotonically decreasing m→∞ sequence εcons(m), such that εcons(m)
−→ 0, and 

D, ES

Dm [F(A(S)) − F

] ≤ εcons(m). A learning rule for
which this holds is denoted as a universally consistent learning rule.”
The above definition of learnability, which requires a uniform rate of all
distributions, is the most appropriate concept to study learnability of
a hypothesis class. It is a direct generalization of “agnostic PAC-
learnability” to “Vapnik’s General Setting of Learning” as studied by
Haussler in 1992. A potential path to learning is a minimization of the
empirical risk “F
S
(h)” over a sample “S”, defined as
“FS ( ) = 1/m ∑ f ( )”
Z,z
“Instance domain and a specific instance.”
H,h
“Hypothesis class and a specific hypothesis.”
f(h,z)
“Loss of hypothesis h on instance z.”
B
“suph,z | (h; z)|”
D
“Underlying distribution on instance domain Z”
S
“Empirical sample z1,...,zm, sampled i.i.d. from D”
m
“Size of empirical sample S”
A(S)
“Learning rule A applied to empirical sample S”
εcons (m) “Rate of consistency for a learning rule”
(h)
“Risk of hypothesis h, Ez

(h; z)]”
“infh

H F(h)”
“Empirical risk of hypothesis h on sample S, 1 ∑z

S f (h; z) m”
“An ERM hypothesisFS(hˆS) = infh

H FS(h) Rate of AERM for a learning rule”
F

FS ( h )
hˆ S
εerm (m)
εstable (m) “Rate of stability for a learning rule”
εgen (m) “Rate of generalization for a learning rule”


The “rule Ais an “Empirical Risk Minimizer” if it can minimize the
empirical risk
“F
S
(A(S)) = F
S
(hˆ
S
) = inf 

Download 1,94 Mb.

Do'stlaringiz bilan baham:
1   ...   73   74   75   76   77   78   79   80   ...   96




Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©hozir.org 2024
ma'muriyatiga murojaat qiling

kiriting | ro'yxatdan o'tish
    Bosh sahifa
юртда тантана
Боғда битган
Бугун юртда
Эшитганлар жилманглар
Эшитмадим деманглар
битган бодомлар
Yangiariq tumani
qitish marakazi
Raqamli texnologiyalar
ilishida muhokamadan
tasdiqqa tavsiya
tavsiya etilgan
iqtisodiyot kafedrasi
steiermarkischen landesregierung
asarlaringizni yuboring
o'zingizning asarlaringizni
Iltimos faqat
faqat o'zingizning
steierm rkischen
landesregierung fachabteilung
rkischen landesregierung
hamshira loyihasi
loyihasi mavsum
faolyatining oqibatlari
asosiy adabiyotlar
fakulteti ahborot
ahborot havfsizligi
havfsizligi kafedrasi
fanidan bo’yicha
fakulteti iqtisodiyot
boshqaruv fakulteti
chiqarishda boshqaruv
ishlab chiqarishda
iqtisodiyot fakultet
multiservis tarmoqlari
fanidan asosiy
Uzbek fanidan
mavzulari potok
asosidagi multiservis
'aliyyil a'ziym
billahil 'aliyyil
illaa billahil
quvvata illaa
falah' deganida
Kompyuter savodxonligi
bo’yicha mustaqil
'alal falah'
Hayya 'alal
'alas soloh
Hayya 'alas
mavsum boyicha


yuklab olish