TR TU NHN TO

Artificial Intelligence

TS. Nguyn nh Thun

Khoa Cng ngh Thng tin
i hc Nha Trang
Email: thuanvinh122@gmail.com

Nha Trang 8-2007

 Noi dung mon hoc

Chng 1: Gii thiu

M u
Lnh vc nghin cu ca AI
ng dng ca AI
Cc vn t ra

Slide 2
 Noi dung mon hoc (tip)

Chng 2:Tm kim trn khng gian trng thi

Bi ton tm kim
Gii thut tng qut
Depth first search (DFS)
Breath first search (BFS)

Chng 3:Tm kim theo Heuristic

Gii thiu v Heuristic
Tm kim theo heuristic
Gii thut Best first search (BFS), Gii thut AT, AKT, A*
Chin lc Minimax, Alpha Beta

Slide 3
 Noi dung mon hoc (tip)

Chng 4:Biu din tri thc

B ba i tng Thuc tnh Gi tr
Cc lut dn
Mng ng ngha
Frame
Logic mnh , Logic v t
Thut gii Vng Ho, Thut gii Robinson

Chng 5: My hc
Cc hnh thc hc
Thut gii Quinland
Hc theo bt nh

Slide 4
 Thc hnh &Ti liu tham kho

Thc hnh Prolog / C++ / Pascal

Cc gii thut tm kim
Biu din tri thc
Bi tp ln

Ti liu tham kho
Bi ging Tr tu nhn to TS Nguyn nh Thun
Gio trnh Tr tu nhn to - GS Hong Kim HQGTPHCM
Tr tu nhn toPGS Nguyn Thanh ThyH Bch Khoa HNi
Artificial Inteligent George F. Luget & Cilliam A. Stubblefied

Slide 5
Chng 1: GII THIEU

TS. Nguyn nh Thun

Khoa Cng ngh Thng tin
i hc Nha Trang
Email: thuanvinh122@gmail.com
 1.1 M u

Tr tu l g:
Theo t in Bch khoa ton th Webster:

Tr tu l kh nng:
Phn ng mt cch thch hp li nhng tnh hung
mi thng qua iu chnh hnh vi mt cch thch
hp.
Hiu r mi lin h gia cc s kin ca th gii bn
ngoi nhm a ra nhng hnh vi ph hp t
c mc ch.

Slide 7
 S Thng Minh

Khi nim v tnh thng minh ca mt i

tng thng biu hin qua cc hot ng:

S hiu bit v nhn thc c tri thc

S l lun to ra tri thc mi da trn tri thc
c

Hnh ng theo kt qu ca cc l lun

K nng (Skill)

TRI THC ???

Slide 8
 Tri thc (Knowledge)

Tri thc l nhng thng tin cha ng 2 thnh phn
Cc khi nim:

Cc khi nim c bn: l cc khi nim mang tnh quy c

Cc khi nim pht trin: c hnh thnh t cc khc nim c bn
thnh cc khi nim phc hp phc tp hn.
Cc phng php nhn thc:

Cc qui lut, cc th tc

Phng php suy din, l lun,..

Tri thc l iu kin tin quyt ca cc hnh x thng minh hay
S thng minh

Tri thc c c qua s thu thp tri thc v sn sinh tri thc

Qu trnh thu thp v sn sinh tri thc l hai qu trnh song song v
ni tip vi nhau khng bao gi chm dt trong mt thc th
Thng Minh
Slide 9
 Tri thc Thu thp v sn sinh

Thu thp tri thc:

Tri thc c thu thp t thng tin, l kt qu ca mt qu
trnh thu nhn d liu, x l v lu tr. Thng thng qu
trnh thu thp tri thc gm cc bc sau:

Xc nh lnh vc/phm vi tri thc cn quan tm

Thu thp d liu lin quan di dng cc trng hp c th.

H thng ha, rt ra nhng thng tin tng qut, i din cho cc
trng hp bit Tng qut ha.

Xem xt v gi li nhng thng tin lin quan n vn cn quan
tm , ta c cc tri thc v vn .

Sn sinh tri thc:
Tri thc sau khi c thu thp s c a vo mng tri thc c.
Trn c s thc hin cc lin kt, suy din, kim chng sn sinh ra
cc tri thc mi.
Slide 10
 Tri thc Tri thc siu cp

Tr thc siu cp (meta knowledge) hay Tri thc v

Tri thc
L cc tri thc dng :
nh gi tri thc khc
nh gi kt qu ca qu trnh suy din
Kim chng cc tri thc mi

Phng tin truyn tri thc: ngn ng t nhin

Slide 11
 Hanh x thong minh Ket luan

Hnh x thng minh khng n thun l cc hnh ng nh l kt

qu ca qu trnh thu thp tri thc v suy lun trn tri thc.

Hnh x thng minh cn bao hm
S tng tc vi mi trng nhn cc phn hi
S tip nhn cc phn hi iu chnh hnh ng - Skill
S tip nhn cc phn hi hiu chnh v cp nht tri thc

Tnh cht thng minh ca mt i tng l s tng hp ca c 3
yu t: thu thp tri thc, suy lun v hnh x ca i tng trn tri
thc thu thp c. Chng ha quyn vo nhau thnh mt th
thng nht S Thng Minh

Khng th nh gi ring l bt k mt kha cnh no ni v
tnh thng minh.
 THNG MINH CN TRI THC

Slide 12
 1.2 i tng nghien cu cua AI

AI l lnh vc ca Cng ngh thng tin, c chc nng nghin

cu v to ra cc chng trnh m phng hot ng t duy ca
con ngi.

Tr tu nhn to nhm to ra My ngi?
Mc tiu

Xy dng l thuyt v thng minh gii thch cc hot ng
thng minh

Tm hiu c ch s thng minh ca con ngi
C ch lu tr tri thc
C ch khai thc tri thc

Xy dng c ch hin thc s thng minh

p dng cc hiu bit ny vo cc my mc phc v con
ngi.
Slide 13
 1.2 i tng nghien cu cua AI(tip)

AI l ngnh nghin cu v cch hnh x thng minh

(intellgent behaviour) bao gm: thu thp, lu tr tri
thc, suy lun, hot ng v k nng.

i tng nghin cu l cc hnh x thng minh
ch khng phi l s thng minh.

Gii quyt bi ton bng AI l tm cch biu din tri
thc, tm cch vn dng tri thc gii quyt vn
v tm cch b sung tri thc bng cch pht hin tri
thc t nhng thng tin sn c (my hc)

Slide 14
 1.3 Lch s pht trin ca AI :
Giai on c in

Giai on c in (1950 1965)

C 2 k thut tm kim c bn:
K thut generate and test : ch tm c 1 p n/ cha
chc ti u.
K thut Exhaustive search (vt cn): Tm tt c cc
nghim, chn la phng n tt nht.

(Bng n t hp mn vi m>=10)

Slide 15
 Lch s pht trin ca AI :
Giai on vin vng

Giai on vin vng (1965 1975)

y l giai on pht trin vi tham vng lm cho my hiu c
con ngi qua ngn ng t nhin.
Cc cng trnh nghin cu tp trung vo vic biu din tri thc v
phng thc giao tip gia ngi v my bng ngn ng t nhin.
Kt qu khng my kh quan nhng cng tm ra c cc phng
thc biu din tri thc vn cn c dng n ngy nay tuy cha
tht tt nh:

Semantic Network (mng ng ngha)

Conceptial graph ( th khi nim)

Frame (khung)
Vp phi tr ngi v nng lc

Script (kch bn) ca my tnh

Slide 16
 Lch s pht trin ca AI :
Giai on hin i

Giai on hin i (t 1975)

Xc nh li mc tiu mang tnh thc tin hn ca AI:

Tm ra li gii tt nht trong khong thi gian chp nhn c.

Khng cu ton tm ra li gii ti u
Tinh thn HEURISTIC ra i v c p dng mnh m khc
phc bng n t hp.
Khng nh vai tr ca tri thc ng thi xc nh 2 tr ngi ln l
biu din tri thc v bng n t hp.
Nu cao vai tr ca Heuristic nhng cng khng nh tnh kh khn
trong nh gi heuristic.
Pht trin ng dng mnh m: H chuyn gia,
Better than nothing H chun on,..

Slide 17
 1.4 Cc lnh vc ng dng

Game Playing: Tm kim / Heuristic

Automatic reasoning & Theorem proving: Tm kim / Heuristic

Expert System: l hng pht trin mnh m nht v c gi tr ng
dng cao nht.

Planning & Robotic: cc h thng d bo, t ng ha

Machine learning: Trang b kh nng hc tp gii quyt vn kho
tri thc:
Supervised : Kim sot c tri thc hc c. Khng tm ra ci mi.
UnSupervised:T hc, khng kim sot. C th to ra tri thc mi nhng cng
nguy him v c th hc nhng iu khng mong mun.

Slide 18
 1.4 Cc lnh vc ng dng(tip)

Natural Language Understanding & Semantic modelling:

Khng c pht trin mnh do mc phc tp ca bi
ton c v tri thc & kh nng suy lun.

Modeling Human perfromance: Nghin cu c ch t chc
tr tu ca con ngi p dng cho my.

Language and Environment for AI:Pht trin cng c v mi
trng xy dng cc ng dng AI.

Neural network / Parallel Distributed processing: gii quyt
vn nng lc tnh ton v tc tnh ton bng k thut
song song v m phng mng thn kinh ca con ngi.

Slide 19
 ng dung AI

M hnh ng dng AI hin ti:

AI = Presentation & Search

Mc d mc tiu ti thng ca ngnh TTNT l xy dng mt chic my c
nng lc t duy tng t nh con ngi nhng kh nng hin ti ca tt c
cc sn phm TTNT vn cn rt khim tn so vi mc tiu ra. Tuy
vy, ngnh khoa hc mi m ny vn ang tin b mi ngy v ang t ra
ngy cng hu dng trong mt s cng vic i hi tr thng minh ca con
ngi. Hnh nh sau s gip bn hnh dung c tnh hnh ca ngnh tr tu
nhn to.

Slide 20
 Cc bi ton

Xt cc bi ton sau:
1. i tin (Vt cn v Heuristic)
2. Tm kim chiu rng v su
3. Tic tac toe.
4. ong du.
5. Bi ton TSP
6. 8 puzzle.
7. C vua
8. C tng
9. Ngi nng dn qua sng.
10. Con th v con co
11. Con kh v ni chui
Slide 21
Chng 2: TM KIM TRN KHNG
GIAN TRNG THI
(State Space Search)

TS. Nguyn nh Thun

Khoa Cng ngh Thng tin
i hc Nha Trang
Email: thuanvinh122@gmail.com
 Bi ton tm kim

Tm kim ci g?

Biu din v tm kim l k thut ph bin gii cc bi
ton trong lnh vc AI

Cc vn kh khn trong tm kim vi cc bi ton
AI
c t vn phc tp
Khng gian tm kim ln
c tnh i tng tm kim thay i
p ng thi gian thc
Meta knowledge v kt qu ti u

Kh khn v k thut
Slide 23
 Cu trc chung ca bi ton tm kim

Mt cch chung nht, nhiu vn -bi ton phc tp u

c dng "tm ng i trong th" hay ni mt cch
hnh thc hn l "xut pht t mt nh ca mt th,
tm ng i hiu qu nht n mt nh no ".

Mt pht biu khc thng gp ca dng bi ton ny l:
Cho trc hai trng thi T0 v TG hy xy dng chui trng thi
T0, T1, T2, ..., Tn-1, Tn = TG sao cho :

tha mn mt iu kin cho trc (thng l nh nht).

Slide 24
 2.2 Gii thut tng qut

K hiu:
s nh xut pht
g: nh ch
n: nh ang xt
(n): tp cc nh c th i trc tip t nh n
Open: tp cc nh c th xt bc k tip
Close: tp cc nh xt

Slide 25
 2.2 Gii thut tng qut (tip)

Begin
Open := {s};
Close := ;
While (Open <> ) do
begin
n:=Retrieve(Open);
if (n=g) then Return True;
Open := Open (n); // ((n) Close)
Close := Close {n};
end;
Return False;
End;
Slide 26
 V d:

Xt graph sau:
A s = A l nh bt u
g= G l nh ch
B C D

E F G

H I J

Slide 27
 2.3 Breath First Search V d

Ln lp n Open Close

Xt graph sau: (n)
0 {A}
A 1 A {B, C, D} {B, C, D} {A}
2 B {E, F} {C, D, E, F} {A, B}
3 C {F, G} {D, E, F, G} {A, B, C}
B C D4 D {E, F, G} {A, B, C, D}
5 E {H, I} {F, G, H, I} {A, B, C, D, E}
6 F {J} {G, H, I, J} {A, B, C, D, E, F}
E F G 7 G
True

H I J

Slide 28
 2.3 Breath First Search V d 1

Ln lp n Open Close

Xt graph sau:A->U (n)
0 {A}
A 1 A {B, C, D} {B,C,D} {A}
2 B {E, F} {C,D, E,F} {A, B}
3 C {F, G} {D,E, F,G} {A, B, C}
B C D4 D {E, F, G} {A, B, C, D}
5 E {H, I} {F, G, H, I} {A, B, C, D, E}
6 F {J} {G, H, I, J} {A, B, C, D, E, F}
E F G 7 G {H, I, J} {A, B, C, D, E, F,G}
8 H {I, J} {A,B,C, D, E, F,G,H}

9 I {J} {A,B,C, D, E,
10 J F,G,H,I}
H I J {A,B,C, D, E,
FALSE F,G,H,I,J}

Slide 29
 V d:

Xt graph sau:
A

B C D

E F G

H I J

Slide 30
 2.4 Depth First Search V d

Xt graph sau: Ln lp n (n) Open Close

0 {A}
A 1 A {B, C, D} {B, C, D} {A}
2 B {E, F} {E, F, C, D} {A, B}
B C 3 DE {H, I} {H, I, F, C, D} {A, B, E}
4 H {I, F, C, D} {A, B, E, H}
5 I {F, C, D} {A, B, E, H, I}
E F G 6 F {J} {J, C, D} {A, B, E, H, I, F}
7 J {C, D} {A, B, E, H, I, F,J}
8 C {F, G} {G, D} {A,B,E,H,I, F,J,C}
9 G True
H I J

Slide 31
 Breath First vs Depth First

Breath First: open c t chc dng FIFO

Depth First: open c t chc dng LIFO

Hiu qu
Breath First lun tm ra nghim c s cung nh nht
Depth First thng cho kt qu nhanh hn.

Kt qu
BFS, DFS chc chn tm ra kt qu nu c.

Bng n t hp l kh khn ln nht cho cc gii thut
ny.
Gii Php cho bng n t hp??

Slide 32
 Tm Kim Rng
1. Open = [A]; closed = []
2. Open = [B,C,D];
closed = [A]
2. Open = [C,D,E,F];
closed = [B,A]
3. Open = [D,E,F,G,H]; closed = [C,B,A]
4. Open = [E,F,G,H,I,J]; closed = [D,C,B,A]
5. Open = [F,G,H,I,J,K,L];closed = [E,D,C,B,A]
6. Open = [G,H,I,J,K,L,M];(v L c trong open);

closed = [F,E,D,C,B,A]

Slide 33
Tm kim Su

1. Open = [A]; closed = []

2. Open = [B,C,D]; closed = [A]
3. Open = [E,F,C,D];closed = [B,A]
4. Open = [K,L,F,C,D];
closed = [E,B,A]
5. Open = [S,L,F,C,D];
closed = [K,E,B,A]
6. Open = [L,F,C,D];
closed = [S,K,E,B,A]
7. Open = [T,F,C,D];
closed = [L,S,K,E,B,A]
8. Open = [F,C,D];
closed = [T,L,S,K,E,B,A]
9.

Slide 34
 Depth first search c gii hn

Depth first search c kh nng lp v tn do cc trng

thi con sinh ra lin tc. su tng v tn.

Khc phc bng cch gii hn su ca gii thut.

Su bao nhiu th va?

Chin lc gii hn:
C nh mt su MAX, nh cc danh th chi c tnh
trc c s nc nht nh
Theo cu hnh resource ca my tnh
Meta knowledge trong vic nh gii hn su.

Gii hn su => co hp khng gian trng thi => c
th mt nghim.
Slide 35
Chng 3: HEURISTIC SEARCH

TS. Nguyn nh Thun

Khoa Cng ngh Thng tin
i hc Nha Trang
Email: thuanvinh122@gmail.com
 3.1 Gii thiu v Heuristic

Heuristic l g?
Heuristic l nhng tri thc c rt ta t nhng kinh
nghim, trc gic ca con ngi.
Heuristic c th l nhng tri thc ng hay sai.
Heuristic l nhng meta knowledge v thng ng.

Heuristic dng lm g?
Trong nhng bi ton tm kim trn khng gian trng thi, c
2 trng hp cn n heuristic:

Vn c th khng c nghim chnh xc do cc mnh khng pht
biu cht ch hay thiu d liu khng nh kt qu.

Vn c nghim chnh xc nhng ph tn tnh ton tm ra nghim
l qu ln (h qu ca bng n t hp)
Heuristic gip tm kim t kt qu vi chi ph thp hn
Slide 37
 Heuristic (tip)

Thut gii Heuristic l mt s m rng khi

nim thut ton. N th hin cch gii bi ton
vi cc c tnh sau:
Thng tm c li gii tt (nhng khng chc l
li gii tt nht)
Gii bi ton theo thut gii Heuristic thng d
dng v nhanh chng a ra kt qu hn so vi gii
thut ti u, v vy chi ph thp hn.
Thut gii Heuristic thng th hin kh t nhin,
gn gi vi cch suy ngh v hnh ng ca con
ngi.
Slide 38
 Heuristic (tip)

C nhiu phng php xy dng mt thut gii Heuristic, trong

ngi ta thng da vo mt s nguyn l c bn nh sau:

Nguyn l vt cn thng minh: Trong mt bi ton tm kim no , khi khng
gian tm kim ln, ta thng tm cch gii hn li khng gian tm kim hoc
thc hin mt kiu d tm c bit da vo c th ca bi ton nhanh chng
tm ra mc tiu.

Nguyn l tham lam (Greedy): Ly tiu chun ti u (trn phm vi ton cc)
ca bi ton lm tiu chun chn la hnh ng cho phm vi cc b ca tng
bc (hay tng giai on) trong qu trnh tm kim li gii.

Nguyn l th t: Thc hin hnh ng da trn mt cu trc th t hp l ca
khng gian kho st nhm nhanh chng t c mt li gii tt.

Hm Heuristic: Trong vic xy dng cc thut gii Heuristic, ngi ta thng
dng cc hm Heuristic. l cc hm nh gi th, gi tr ca hm ph thuc
vo trng thi hin ti ca bi ton ti mi bc gii. Nh gi tr ny, ta c th
chn c cch hnh ng tng i hp l trong tng bc ca thut gii.

Slide 39
 Heuristic Greedy

Bi ton i tin: i s tin n thnh cc loi tin cho

trc sao cho s t l t nht

Bi ton hnh trnh ngn nht (TSP): Hy tm mt hnh
trnh cho mt ngi giao hng i qua n im khc nhau,
mi im i qua mt ln v tr v im xut pht sao cho
tng chiu di on ng cn i l ngn nht. Gi s
rng c con ng ni trc tip t gia hai im bt k.
Vt cn: (n-1)! (Vi n ln ???)
Greedy 1: Mi bc chn i j sao cho j gn i nht trong nhng
nh ni vi i cn li
Greedy 2: Mi bc chn i j sao cho i gn j nht trong nhng
nh ni vi j cn li
Slide 40
 V d: TSP vi n=8

1 2 3 4 5 6 7 8
1 0 730 640 840 800 430 380 1010
2 730 0 710 1040 500 300 540 470
3 640 710 0 1420 1050 600 920 1160
4 840 1040 1420 0 740 950 570 900
5 800 500 1050 740 0 520 460 200
6 430 300 600 950 520 0 390 690
7 380 540 920 570 460 390 0 660
8 1010 470 1160 900 200 690 660 0

Slide 41
 V d: TSP vi n=8

*Vi Greedy 1:
1 7 6 2 8 5 4 3 1
Tng chi ph: 4540
*Vi Greedy 2:
1 7 4 5 8 2 6 3 1
Tng chi ph: 3900

Bi ton 3: Bi ton t mu bn

Slide 42
 Heuristic (tt)

Heuristic dng nh th no trong tm kim?

Tm kim trn khng gian trng thi theo chiu no? Su
hay rng?
Tm theo Heuristic : Heuristic nh hng qu trnh tm
kim theo hng m n cho rng kh nng t ti nghim
l cao nht. Khng su cng khng rng

Kt qu ca tm kim vi Heuristic
Vic tm kim theo nh hng ca heuristic c kt qu tt
hay xu ty theo heuristic ng hay sai.
Heuristic c kh nng b st nghim
Heuristic cng tt cng dn n kt qu nhanh v tt.
Lm sao tm c Heuristic tt???
Slide 43
 3.2 Tm kim ti u (Best First Search)

OPEN : tp cha cc trng thi c sinh ra nhng cha c xt n (v ta

chn mt trng thi khc). Thc ra, OPEN l mt loi hng i u tin
(priority queue) m trong , phn t c u tin cao nht l phn t tt nht.
CLOSE : tp cha cc trng thi c xt n. Chng ta cn lu tr nhng
trng thi ny trong b nh phng trng hp khi mt trng thi mi c
to ra li trng vi mt trng thi m ta xt n trc .
Thut gii BEST-FIRST SEARCH

1. t OPEN cha trng thi khi u.

2. Cho n khi tm c trng thi ch hoc khng cn nt no trong OPEN, thc
hin :
2.a. Chn trng thi tt nht (Tmax) trong OPEN (v xa Tmax khi OPEN)
2.b. Nu Tmax l trng thi kt thc th thot.
2.c. Ngc li, to ra cc trng thi k tip Tk c th c t trng thi Tmax.
i vi mi trng thi k tip Tk thc hin : Tnh f(Tk); Thm Tk vo OPEN

Slide 44
 3.2 Tm kim ti u (tip)
Thut gii BEST-FIRST SEARCH

Begin

Open := {s};
Begin
Close := ;
open:={s};

While (Open <> ) do
While (open<> ) do
begin
begin
n:= Retrieve(Open) //Chn trng thi tt nht t
Open. n:=Retrieve(Open);
if (n=g) then return True
if (n=g) then Return True;
else begin
Open := Open (n); // ((n) Close)
To (n)

Close := Close {n};
for mi nt con m ca (n) do
Gn gi tr chi ph cho m
end;
Open:=Open{m};
Return False;
end;
End;
Return False;

End;

Slide 45
 3.2 Tm kim ti u (tip)

- BFS kh n gin. Tuy vy, trn thc t, cng nh tm kim

chiu su v chiu rng, him khi ta dng BFS mt cch trc
tip. Thng thng, ngi ta thng dng cc phin bn ca
BFS l AT, AKT v A*
Thng tin v qu kh v tng lai
- Thng thng, trong cc phng n tm kim theo kiu BFS,
tt f ca mt trng thi c tnh da theo 2 hai gi tr m ta
gi l l g v h. h chng ta bit, l mt c lng v chi
ph t trng thi hin hnh cho n trng thi ch (thng tin
tng lai). Cn g l "chiu di qung ng" i t trng thi
ban u cho n trng thi hin ti (thng tin qu kh). Lu
rng g l chi ph thc s (khng phi chi ph c lng).

Slide 46
 3.3 Thut gii AT

Phn bit khi nim g v h

Slide 47
 3.3 Thut gii AT

Thut gii AT l mt phng php tm kim theo kiu BFS vi

tt ca nt l gi tr hm g tng chiu di con ng i
t trng thi bt u n trng thi hin ti.
Begin
open:={s};
While (open<> ) do
begin
n:= Retrieve(Open) //Chn n sao cho g(n) nh nht t Open.
if (n=g) then return True
else begin
To (n)
for mi nt con m ca (n) do
if (mOpen) then
Begin
g(m):=g(n)+Cost(n,m)
Open:=Open{m};
end
else So snh g(m) va gNew (m) v cp nht
end;
Return False; Slide 48
End;
 3.3 Thut gii CMS (Cost Minimazation Search)

Thut gii CMS l mt phng php tm kim theo kiu BFS vi tt ca

nt l gi tr hm g v b sung tp Close: tp nh xt).
Begin
open:={s}; close :=
While (open<> ) do
begin
n:= Retrieve(Open) //Chn n sao cho g(n) nh nht t Open.
if (n=g) then return True
else begin
To (n)
for mi nt con m ca (n) do
if (mOpen) and (m Close) then
Begin
g(m):=g(n)+Cost(n,m)
Open:=Open{m};
end
else So snh g(m) va gNew (m) v cp nht
close = close {n}
end;
Return False;
Slide 49
End;
 V d:

Xt graph sau:
A 30
20 35
B C D
40 45 15 10
E F G
30 25 10 20
H I J

s = A l nh bt u
g= J l nh ch
Slide 50
 V d:

Ln lp n (n) Open

Xt graphSau
Trc sau:
* A 0 {(A,0)}

A B 1 A {B,C,D} {(B,20), (C,35), (D,30)}

A C
B {E,F} (C,35), (D,30),(E,60),(F,65)
A D
B E D (C,35),(E,60),(F,65)

C F C {F,G} (E,60),(F,50),(G,45)
C G
F J G {J} (E,60),(F,50),(J,65)

s = A l nh bt u F (J} (E,60),(J,60)

g= J l nh ch J
Slide 51
 3.4 Thut gii AKT

(Algorithm for Knowlegeable Tree Search)

Thut gii AKT m rng AT bng cch s dng thm thng tin c lng h. tt
ca mt trng thi f l tng ca hai hm g v h.
Begin
open:={s};
While (open<> ) do
begin
n:= Retrieve(Open) //Chn n sao cho f(n) nh nht t Open.
if (n=g) then return True
else begin
To (n)
for mi nt con m ca (n) do
Begin
g(m):=g(n)+Cost(n,m)
f(m):= g(m)+h(m);
Open:=Open{m};
end;
end;
Return False;
End; Slide 52
 3.5 Thut gii A*

Thut gii A*
A* l mt phin bn c bit ca AKT p dng cho trng hp th.
Thut gii A* c s dng thm tp hp CLOSE lu tr nhng
trng hp c xt n. A* m rng AKT bng cch b sung cch
gii quyt trng hp khi "m" mt nt m nt ny c sn trong
OPEN hoc CLOSE.

Slide 53
 3.5 Thut gii A* (tip)

Begin
open:={s}; close:=; m Open:
While (open<> ) do if n c m bng mt path ngn hn
begin then Cp nht li m trong Open.
n:= Retrieve(Open) //sao cho f(n) min. m Close
if (n=g) then return path t s n g if n c m bng mt path ngn hn
then begin
else begin
Close:=Close-{m}
To (n)
Open:=Open{m}
for mi nt con m ca (n) do end;
case m of end; /*end case*/
m Open v m Close: Close:=Close{n}
begin end; / while/
Gn gi tr heuristic cho m return false;
Open:=Open{m}; End;
end;

Slide 54
 Hm lng gi Heuristic

Hm lng gi Heuristic l hm c lng ph tn i t trng thi

hin ti n trng thi goal.

C s xc nh hm lng gi l da vo tri thc/kinh nghim thu
thp c.

Hm lng gi cho kt qu ng (gn thc th) hay sai (xa gi tr
thc) s dn n kt qu tm c tt hay xu.

Khng c chun mc cho vic nh gi mt hm lng gi
Heuristic. L do:
Khng c cu trc chung cho hm lng gi
Tnh ng/sai thay i lin tc theo tng vn c th
Tnh ng/sai thay i theo tng tnh hung c th trong mt vn

C th dng nhiu hm lng gi khc nhau theo tnh hung 
cn hm lng gi v cc hm lng gi.
Slide 55
 Tr 8 hay 15

Trng thi ban u Trng thi ch

Tr 11 14 4 7 1 2 3 4
15 10 6 5 12 13 14 5
1 2 13 15 11 15 6
9 12 8 3 10 9 8 7

Tr
2 8 1 2 3
8 3 5 7 8 4
6 2 1 7 6 5

Cn biu din KGTT cho bi ton ny nh th no?

C 3 Tm kim khng gian trng thi Slide 56

 Thut gii A* V d

l Xt bi ton 8 pussle 2 8 3
vi goal l: 1 6 4
5 6
1 2 3 7 5

8 4
7 6 5 2 8 3
1 4 3 4
7 6 5
Heuristic 1: Tng s
ming sai v tr 2 8 3
Heuristic 2: Tng 1 6 4 5 6
khong cch sai v tr 7 5
ca tng ming.
Vic chn la hm Heuristic l kh khn v c
ngha quyt nh i vi tc ca gii thut

Slide 57
Hm lng gi Heuristic Cu trc

Xt li hot ng ca gii thut Best First Search:

Khi c 2 nt cng c gi tr k vng t n mc tiu bng nhau th nt c
path t nt bt u n nt ngn hn s c chn trc nh vy nt ny
c gi tr Heuristic tt hn.
Hay ni cch khc hm lng gi Heuristic cho nt gn start hn l tt hn
nu k vng n goal l bng nhau.
Vy chn nt no nu k vng ca 2 nt khc nhau? Nt k vng tt hn
nhng xa start hay nt k vng xu hn nhng gn root
Hm lng gi bao gm c 2 v c cu trc:
F(n) := G(n) + H(n)
G(n): ph tn thc t root n n
H(n): ph tn c lung heuristic t n n goal.

Slide 58
 Thut gii A* V d

l Xt v d l bi ton 8 puzzle vi:

2 8 3 1 2 3
1 6 4 8 4
7 5 7 6 5

Bt u ch

Hm lng gi: F(n) = G(n) + H(n)

Vi G(n): s ln chuyn v tr thc hin
H(n): S ming nm sai v tr
Nt X c gi tr heuristic tt hn nt Y nu F(x) < F(y).

Ta c hot ng ca gii thut Best First search trn nh hnh sau:

Slide 59
 3.5 Thut gii A* (tip)

Begin
open:={s}; close:=; m Open:
While (open<> ) do if n c m bng mt path ngn hn
begin then Cp nht li m trong Open.
n:= Retrieve(Open) //sao cho f(n) min. m Close
if (n=g) then return path t s n g if n c m bng mt path ngn hn
then begin
else begin
Close:=Close-{m}
To (n)
Open:=Open{m}
for mi nt con m ca (n) do end;
case m of end; /*end case*/
m Open v m Close: Close:=Close{n}
begin end; / while/
Gn gi tr heuristic cho m return false;
Open:=Open{m}; End;
end;

Slide 60
 V d
1 2 8 3
State A
1 6 4
7 5 F(a) =0+4=4

x 2 8 3
State B 2 2 8 3
State C x 2 8 3
State D
1 6 4 1 4 1 6 4
7 5 F(b) =1+5=6 7 6 5 F(c) =1+3=4 7 5 F(D) =1+5=6

3 2 8 3
State E 4 2 3
State F x 2 8 3
State G 2 8 3
1 4 1 8 4 1 4 1 6 4
7 6 5 F(e) =2+3=5 7 6 5 F(f) =2+3=5 7 6 5 F(g) =2+4=6 7 5

x 8 3
State H x 2 8 3
State I
2 1 4 7 1 4 2 8 3
7 6 5 F(h) =3+3=6 6 5 F(i) =3+4=7 1 4
7 6 5
Slide 61
 V d

4 2 3
State F
1 8 4
7 6 5 F(f) =2+3=5

5 2 3
State J x 2 3
State K y 2 8 3
State Close
1 8 4 1 8 4 1 4
7 6 5 F(j) =3+2=5 7 6 5 F(k) =3+4=7 7 6 5

6 1 2 3
State L y 2 3
8 4
Close
1 8 4
7 6 5 F(l) =4+1=5 7 6 5

y 2 3
State Close 7 1 2 3
State M x 1 2 3
State N
1 8 4 8 4 7 8 4
7 6 5 7 6 5 F(m) =5+0=5 6 5 F(n) =5+1=7

Slide 62
C 3 Tm kim khng gian trng thi

The 8-puzzle searched by a production system with

loop detection and depth bound 5

Tr chi 8-puzzle
Slide 63
Hot ng theo gii thut A*

Ln n Open Close
0 {A4} {}
1 A4 {C4,B6,D6} {A4}
2 C4 {E5,F5,G6,B6,D6} {A4,C4}
3 E5 {F5,H6,G6,B6,D6,I7} {A4,C4,E5}
4 F5 {J5,H6,G6,B6,D6,K7,I7} {A4,C4,E5,F5}
5 J5 {L5,H6,G6,B6,D6,K7,I7} {A4,C4,E5,F5,J5}
6 l5 {M5,H6,G6,B6,D6,K7,I7,N7} {A4,C4,E5,F5,J5,L5}
7 m5

Slide 64
nh gi gii thut Heuristic

Admissibility Tnh chp nhn
Mt gii thut Best first search vi hm nh gi
F(n) = G(n) + H(n) vi

N : Trng thi bt k

G(n) : Ph tn i t nt bt u n nt n

H(n) : Ph tn c lng heuristic i t nt n n goal
c gi l gii thut A

Mt gii thut tm kim c xem l admissible nu i
vi mt th bt k n lun dng path nghim tt nht
(nu c).

Gii thut A*: L gii thut A vi hm heuristic H(n)lun
lun gi tr thc i t n n goal.

Gii thut A* l admissible Slide 65
nh gi gii thut Heuristic

Monotonicity n iu
Mt hm heuristic H(n) c gi l monotone (n iu) nu:
ni, nj : nj l nt con chu ca ni ta c
H(ni)-H(nj) ph tn tht i t ni n nj
nh gi heuristic ca ch l 0 : H(goal) = 0.

Gii thut A c hm H(n) monotone l gii thut A* v
Admissible

Informedness

Xt 2 hm heuristic H1(n) v H2(n) nu ta c H1(n)
H2(n) vi mi trng thi n th H2(n) c cho l
informed hn H1(n).
Slide 66
Heuristic trong tr chi i khng

Gii thut minimax:
Hai u th trong tr chi c gi l MIN v MAX.
Mi nt l c gi tr:

1 nu l MAX thng,

0 nu l MIN thng.
Minimax s truyn cc gi tr ny ln cao dn trn th, qua cc
nt cha m k tip theo cc lut sau:

Nu trng thi cha m l MAX, gn cho n gi tr ln nht c trong cc trng
thi con.

Nu trng thi b, m l MIN, gn cho n gi tr nh nht c trong cc trng
thi con.

C 4 Tm kim Heuristic Slide 67

 Hy p dng GT Minimax vo Tr Chi NIM

C 4 Tm kim Heuristic Slide 68

 Minimax vi su lp c nh

Minimax i vi mt KGTT gi nh.

Cc nt l c gn cc gi tr heuristic

Cn gi tr ti cc nt trong l cc gi tr nhn c da trn
gii thut Minimax

C 4 Tm kim Heuristic Slide 69

 Heuristic trong tr chi tic-tac-toe

Hm Heuristic: E(n) = M(n) O(n)

Trong : M(n) l tng s ng thng c th ca ti
O(n) l tng s ng thng c th ca i th
E(n) l tr s nh gi tng cng cho trng thi n
C 4 Tm kim Heuristic Slide 70
 Minimax 2 lp trong tic-tac-toe

Trch t Nilsson (1971).

C 4 Tm kim Heuristic Slide 71
 Gii thut ct ta -

Tm kim theo kiu depth-first.

Nt MAX c 1 gi tr (lun tng)

Nt MIN c 1 gi tr (lun gim)

TK c th kt thc di bt k:
Nt MIN no c ca bt k nt cha MAX no.
Nt MAX no c ca bt k nt cha MIN no.

Gii thut ct ta - th hin mi quan h gia cc
nt lp n v n+2, m ti ton b cy c gc ti
lp n+1 c th ct b.

C 4 Tm kim Heuristic Slide 72

 Ct ta
S=
MAX

MIN A= Z
- cut

=z
z

C 4 Tm kim Heuristic Slide 73

 Ct ta

MIN S =

MAX A= Z
- cut

=z
z

C 4 Tm kim Heuristic Slide 74

 GT Ct Ta -
p dng cho KGTT gi nh

Cc nt khng c gi tr l
cc nt khng c duyt
qua

C 4 Tm kim Heuristic Slide 75

Chng 4: Biu din v suy lun tri thc

TS. Nguyn nh Thun

Khoa Cng ngh Thng tin
i hc Nha Trang
Email: thuanvinh122@gmail.com
4.1. M u

tri thc, lnh vc v biu din tri thc.
4.2. Cc loi tri thc: c chia thnh 5 loi
1. Tri thc th tc: m t cch thc gii quyt mt vn . Loi
tri thc ny a ra gii php thc hin mt cng vic no
. Cc dng tri thc th tc tiu biu thng l cc lut,
chin lc, lch trnh v th tc.
2. Tri thc khai bo: cho bit mt vn c thy nh th no.
Loi tri thc ny bao gm cc pht biu n gin, di dng
cc khng nh logic ng hoc sai. Tri thc khai bo cng c
th l mt danh sch cc khng nh nhm m t y hn
v i tng hay mt khi nim no .

Slide 77
 4.2. Cc loi tri thc (tip)
3. Siu tri thc: m t tri thc v tri thc. Loi tri thc ny gip la
chn tri thc thch hp nht trong s cc tri thc khi gii quyt mt vn .
Cc chuyn gia s dng tri thc ny iu chnh hiu qu gii quyt vn
bng cch hng cc lp lun v min tri thc c kh nng hn c.
4. Tri thc heuristic: m t cc "mo" dn dt tin trnh lp lun.
Tri thc heuristic l tri thc khng bm m hon ton 100% chnh xc v
kt qu gii quyt vn . Cc chuyn gia thng dng cc tri thc khoa
hc nh s kin, lut, sau chuyn chng thnh cc tri thc heuristic
thun tin hn trong vic gii quyt mt s bi ton.
5. Tri thc c cu trc: m t tri thc theo cu trc. Loi tri thc ny
m t m hnh tng quan h thng theo quan im ca chuyn gia, bao
gm khi nim, khi nim con, v cc i tng; din t chc nng v mi
lin h gia cc tri thc da theo cu trc xc nh.

Slide 78
 V d: Hy phn loi cc tri thc sau

1. Nha Trang l thnh ph p.

2. Bn Lan thch c sch.
3. Thut ton tm kim BFS, DFS
4. Thut gii Greedy
5. Mt s cch chiu tng trong vic chi c tng.
6. H thng cc khi nim trong hnh hc.
7. Cch tp vit ch p.
8. Tm tt quyn sch v Tr tu nhn to.
9. Chn loi c phiu mua c phiu.

Slide 79
4.3. CC K THUT BIU DIN TRI THC

4.3.1 B ba i tng-Thuc tnh-Gi tr

4.3.2 Cc lut dn
4.3.3 Mng ng ngha
4.3.4 Frames
4.3.5 Logic

Slide 80
4.3.1 B ba i tng-Thuc tnh-Gi tr

Mt s kin c th c dng xc nhn gi tr ca mt thuc tnh xc

nh ca mt vi i tng. V d, mnh "qu bng mu " xc nhn
"" l gi tr thuc tnh "mu" ca i tng "qu bng". Kiu s kin ny
c gi l b ba i tng-Thuc tnh-Gi tr (O-A-V Object-Attribute-
Value).

Hnh 2.1. Biu din tri thc theo b ba O-A-V

Slide 81
4.3.1 B ba i tng-Thuc tnh-Gi tr (tip)

Trong cc s kin O-A-V, mt i tng c th c nhiu thuc tnh

vi cc kiu gi tr khc nhau. Hn na mt thuc tnh cng c th
c mt hay nhiu gi tr. Chng c gi l cc s kin n tr
(single-valued) hoc a tr (multi-valued). iu ny cho php cc
h tri thc linh ng trong vic biu din cc tri thc cn thit.

Cc s kin khng phi lc no cng bo m l ng hay sai vi
chc chn hon ton. V th, khi xem xt cc s kin, ngi ta
cn s dng thm mt khi nim l tin cy. Phng php truyn
thng qun l thng tin khng chc chn l s dng nhn t chc
chn CF (certainly factor). Khi nim ny bt u t h thng
MYCIN (khong nm 1975), dng tr li cho cc thng tin suy
lun. Khi , trong s kin O-A-V s c thm mt gi tr xc nh
tin cy ca n l CF.

Slide 82
 4.3.2 Cc lut dn

Lut l cu trc tri thc dng lin kt thng tin

bit vi cc thng tin khc gip a ra cc suy lun,
kt lun t nhng thng tin bit.

Trong h thng da trn cc lut, ngi ta thu thp cc
tri thc lnh vc trong mt tp v lu chng trong c
s tri thc ca h thng. H thng dng cc lut ny
cng vi cc thng tin trong b nh gii bi ton.
Vic x l cc lut trong h thng da trn cc lut
c qun l bng mt module gi l b suy din.

Slide 83
4.3.2 Cc lut dn(ti
p)

Cc dng lut c bn: 7 dng

1. Quan h:
IF Bnh in hng
THEN Xe s khng khi ng c
2. Li khuyn:
IF Xe khng khi ng c
THEN i b
3. Hng dn
IF Xe khng khi ng c AND H thng nhin liu tt
THEN Kim tra h thng in

Slide 84
 4.3.2 Cc lut dn(ti
p)

4. Chin lc
IF Xe khng khi ng c
THEN u tin hy kim tra h thng nhin liu, sau kim tra h
thng in
5. Din gii
IF Xe n AND ting gin
THEN ng c hot ng bnh thng
6. Chn on
IF St cao AND hay ho AND Hng
THEN Vim hng
7. Thit k
IF L n AND Da sng
THEN Nn chn Xe Spacy AND Chn mu sng

Slide 85
 4.3.3 Mng ng ngha

Mng ng ngha l mt phng php biu din tri thc dng

th trong nt biu din i tng v cung biu din
quan h gia cc i tng.

Hnh 2.3. "S l Chim" th hin trn mng ng ngha

Slide 86
 4.3.3 Mng ng ngha(ti
p)

Hnh 4.4. Pht trin mng ng ngha

Slide 87
 V d: Gii bi ton tam gic tng qut

C 22 yu t ca tam gic. Nh vy c C322 -1 cch xy dng hay

xc nh mt tam gic.
Theo thng k, c khong 200 cng thc lin quan n cnh v gc 1
tam gic.
gii bi ton ny bng cng c mng ng ngha, s dng khong
200 nh cha cng thc v khong 22 nh cha cc yu t ca
tam gic. Mng ng ngha cho bi ton ny c cu trc nh sau :
nh ca th bao gm hai loi :
nh cha cng thc (k hiu bng hnh ch nht)
nh cha yu t ca tam gic (k hiu bng hnh trn)
Cung : ch ni t nh hnh trn n nh hnh ch nht cho bit yu
t tam gic xut hin trong cng thc no
* Lu : trong mt cng thc lin h gia n yu t ca tam gic, ta gi
nh rng nu bit gi tr ca n-1 yu t th s tnh c gi tr ca
yu t cn li. Chng hn nh trong cng thc tng 3 gc ca tam gic
bng 1800 th khi bit c hai gc, ta s tnh c gc cn li.

Slide 88
 V d: Gii bt tam gic tng qut (tt)

B1 : Kch hot nhng nh hnh trn cho ban

u (nhng yu t c gi tr)
B2 : Lp li bc sau cho n khi kch hot c
tt c nhng nh ng vi nhng yu t cn tnh
hoc khng th kch hot c bt k nh no
na.
Nu mt nh hnh ch nht c cung ni vi n nh
hnh trn m n-1 nh hnh trn c kch hot
th kch hot nh hnh trn cn li (v tnh gi tr
nh cn li ny thng qua cng thc nh hnh
ch nht).

Slide 89
 V d: Gii bt tam gic tng qut (tt)

V d : "Cho hai gc , v chiu di cnh a ca tam

gic. Tnh chiu di ng cao hC".

p=(a+b+c)/2

p Slide 90
 4.3.4 Frame

Hnh 2.6. Cu trc frame

Hnh 2.7. Nhiu mc ca frame m t quan h phc tp hn

Slide 91
 4.3.5 Logic

1. Logic mnh
IF Xe khng khi ng c (A)
AND Khong cch t nh n ch lm l xa (B)
THEN S tr gi lm (C)
B

Lut trn c th biu din li nh sau:A C

2. Logic v t

Logic v t, cng ging nh logic mnh , dng cc k hiu
th hin tri thc. Nhng k hiu ny gm hng s, v t, bin
v hm.

Slide 92
 4.4 SUY DIN D LIU

1. Modus ponens
1. E1
2. E1 E2
3. E2
Nu c tin khc, c dng E2 E3 th E3 c a vo danh sch.
2. Modus tollens
1. E2
2. E1 E2
3. E1

Slide 93
 4.5 Chng minh mnh

Mt trong nhng vn kh quan trng ca logic mnh l chng minh

tnh ng n ca php suy din (a b). y cng chnh l bi ton chng
minh thng gp trong ton hc.

Vi hai php suy lun c bn ca logic mnh (Modus Ponens, Modus
Tollens) cng vi cc php bin i hnh thc, ta cng c th chng minh
c php suy din. Tuy nhin, thao tc bin i hnh thc l rt kh ci t
c trn my tnh. Thm ch iu ny cn kh khn vi c con ngi!

Vi cng c my tnh, c th cho rng ta s d dng chng minh c mi
bi ton bng mt phng php bit l lp bng chn tr . Tuy v l
thuyt, phng php lp bng chn tr lun cho c kt qu cui cng
nhng phc tp ca phng php ny l qu ln, O(2n) vi n l s bin
mnh . Sau y chng ta s nghin cu hai phng php chng minh
mnh vi phc tp ch c O(n).

Slide 94
 4.5 Chng minh mnh

Mt trong nhng vn kh quan trng ca logic mnh l chng minh

tnh ng n ca php suy din (a b). y cng chnh l bi ton chng
minh thng gp trong ton hc.

Vi hai php suy lun c bn ca logic mnh (Modus Ponens, Modus
Tollens) cng vi cc php bin i hnh thc, ta cng c th chng minh
c php suy din. Tuy nhin, thao tc bin i hnh thc l rt kh ci t
c trn my tnh. Thm ch iu ny cn kh khn vi c con ngi!

Vi cng c my tnh, c th cho rng ta s d dng chng minh c mi
bi ton bng mt phng php bit l lp bng chn tr . Tuy v l
thuyt, phng php lp bng chn tr lun cho c kt qu cui cng
nhng phc tp ca phng php ny l qu ln, O(2n) vi n l s bin
mnh . Sau y chng ta s nghin cu hai phng php chng minh
mnh vi phc tp ch c O(n).

Slide 95
 4.5.1 Thut gii Vng Ho

B1 : Pht biu li gi thit v kt lun ca vn theo dng chun sau :

GT1, GT2, ..., GTn KL1, KL2, ..., KLm
Trong cc GTi v KLi l cc mnh c xy dng t cc bin mnh v 3
php ni c bn : , ,
B2 : Chuyn v cc GTi v KLi c dng ph nh.
V d :
p q, (r s), g, p r s, p
p q, p r , p (r s), g, s
B3 : Nu GTi c php th thay th php bng du ","
Nu KLi c php th thay th php bng du ","
V d :
p q , r ( p s) q, s
p, q, r, p s q, s

Slide 96
 4.5.1 Thut gii Vng Ho

B4 : Nu GTi c php th tch thnh hai dng con.

Nu KLi c php th tch thnh hai dng con.
V d :
p, p q q
p, p q
v p, q q
B5 : Mt dng c chng minh nu tn ti chung mt mnh c hai pha.
V d :
p, q q c chng minh
p, p q p p, q
B6 :
a) Nu mt dng khng cn php ni v php ni c hai v v 2 v
khng c chung mt bin mnh th dng khng c chng minh.
b) Mt vn c chng minh nu tt c dng dn xut t dng chun ban u
u c chng minh.
V d: i) p ( p q) q
ii) (p q) ( p r) q r
Slide 97
 4.5.2 Thut gii Robinson

Thut gii ny hot ng da trn phng php chng minh

phn chng v php hp gii Robinson.

Phng php chng minh phn chng:
Chng minh php suy lun (a b) l ng (vi a l gi thit, b l kt
lun).
Phn chng : gi s b sai suy ra b l ng.

Php hp gii Robinson:
i) p ( p q) q
ii) (p q) ( p r) q r
Bi ton c chng minh nu a ng v b ng sinh ra mt mu
thun.

Slide 98
 4.5.2 Thut gii Robinson (tip)

B1 : Pht biu li gi thit v kt lun ca vn di dng chun

nh sau :
GT1, GT2, ..., GTn KL1, KL2, ..., KLm
Trong : GTi v KLj c xy dng t cc bin mnh v cc
php ton : , ,
B2 : Nu GTi c php th thay bng du ","
Nu KLi c php th thay bng du ","
B3 : Bin i dng chun B1 v thnh danh sch mnh nh
sau :
{ GT1, GT2, ..., GTn , KL1, KL2, ..., KLm }
B4 : Nu trong danh sch mnh bc 2 c 2 mnh i ngu
nhau th bi ton c chng minh. Ngc li th chuyn sang
B5. (a v a gi l hai mnh i ngu nhau)

Slide 99
 4.5.2 Thut gii Robinson (tip)

B6 : p dng php hp gii

i) p ( p q) q
ii) (p q) ( p r) q r
B7 : Nu khng xy dng c thm mt mnh mi no v
trong danh sch mnh khng c 2 mnh no i ngu
nhau th vn khng c chng minh.

V d : Chng minh rng

( p q) ( q r) ( r s) ( u s) p u

Slide 100
 Chng 5 My hc

5.1 M U

Cc chng trc tho lun v biu din v suy lun tri
thc. Trong trng hp ny gi nh c sn tri thc v c th
biu din tng minh tri thc.

Tuy vy trong nhiu tinh hung, s khng c sn tri thc nh:
K s tri thc cn thu nhn tri thc t chuyn gia lnh vc.
Cn bit cc lut m t lnh vc c th.
Bi ton khng c biu din tng minh theo lut, s kin hay cc
quan h.

C hai tip cn cho h thng hc:
Hc t k hiu: bao gm vic hnh thc ha, sa cha cc lut tng
minh, s kin v cc quan h.
Hc t d liu s: c p dng cho nhng h thng c m hnh di
dng s lin quan n cc k thut nhm ti u cc tham s. Hc theo
dng s bao gm mng Neural nhn to, thut gii di truyn, bi ton ti
u truyn thng. Cc k thut hc theo s khng to ra CSTT tng
minh.
Slide 101
 5.2 CC HNH THC HC

1. Hc vt: H tip nhn cc khng nh ca cc quyt nh

ng. Khi h to ra mt quyt nh khng ng, h s a ra
cc lut hay quan h ng m h s dng. Hnh thc hc
vt nhm cho php chuyn gia cung cp tri thc theo kiu
tng tc.
2. Hc bng cch ch dn: Thay v a ra mt lut c th cn
p dng vo tnh hung cho trc, h thng s c cung
cp bng cc ch dn tng qut. V d: "gas hu nh b thot
ra t van thay v thot ra t ng dn". H thng phi t mnh
ra cch bin i t tru tng n cc lut kh dng.
3. Hc bng qui np: H thng c cung cp mt tp cc v
d v kt lun c rt ra t tng v d. H lin tc lc cc
lut v quan h nhm x l tng v d mi.
Slide 102
 5.2 CC HNH THC HC (Tip)

4. Hc bng tng t: H thng c cung cp p ng ng cho cc tc

v tng t nhng khng ging nhau. H thng cn lm thch ng p
ng trc nhm to ra mt lut mi c kh nng p dng cho tnh
hung mi.
5. Hc da trn gii thch: H thng phn tch tp cc li gii v d (v kt
qu) nhm n nh kh nng ng hoc sai v to ra cc gii thch dng
hng dn cch gii bi ton trong tng lai.
6. Hc da trn tnh hung: Bt k tnh hung no c h thng lp lun
u c lu tr cng vi kt qu cho d ng hay sai. Khi gp tnh
hng mi, h thng s lm thch nghi hnh vi lu tr vi tnh hung
mi.
7. Khm ph hay hc khng gim st: Thay v c mc tiu tng minh, h
khm ph lin tc tm kim cc mu v quan h trong d liu nhp. Cc
v d v hc khng gim st bao gm gom cm d liu, hc nhn
dng cc c tnh c bn nh cnh t cc im nh.

Slide 103
 V d v CC HNH THC HC

V d:
- H MYCIN
- Mng Neural nhn to
- Thut ton hc Quinland
- Bi ton nhn dng
- My chi c car, c tng

Slide 104
 5.3 THUT GII Quinlan

- L thut ton hc theo quy np dng lut, a mc tiu.

- Do Quinlan a ra nm 1979.
- tng: Chn thuc tnh quan trng nht to cy
quyt nh.
- Thuc tnh quan trng nht l thuc tnh phn loi
Bng quan st thnh cc bng con sao cho t mi bng
con ny d phn tch tm quy lut chung.

Slide 105
 5.3.1 THUT GII A. Quinlan

STT Size Nationality Family Conclusion

1 Small German Single A

2 Large French Single A

3 Large German Single A

4 Small Italian Single B

5 Large German Married B

6 Large Italian Single B

7 Large Italian Married B

8 Small German Married B

Slide 106
Vi mi thuc tnh ca bng quan st:

Xt vector V: c s chiu bng s phn loi

V(Size=Small) = (ASmall, BSmall)
ASmall=S quan st A c Size l Small / Tng s quan st c Size=Small
BSmall= S quan st B c Size l Small / Tng s quan st c Size=Small
V(Size=Small) = (1/3 , 2/3)
V(Size=Large) = (2/5 , 3/5)

Vi thuc tnh Nationality
V(Nat = German)= (2/4 , 2/4)
V(Nat = French) = (1 , 0)
V(Nat = Italian) = (0 , 1)

Thuc tnh Family:
V(Family=Single) = (3/5 ,2/5)
V(Family = Married) = (0, 1)

Slide 107
Vi mi thuc tnh ca bng quan st:

Ch cn xt German STT Size Family Conclusion

Thuc tnh Size:
1 Small Single A
V(Size=Small) = (1/2 , 1/2)
V(Size=Large) = (1/2 , 1/2) 2 Large Single A
Thuc tnh Family:
3 Large Married B
V(Family=Single) = (1, 0)
V(Family=Married) = (0,1) 4 Small Married B

Nationality

Italian French German

Single Married
Slide 108
Vi mi thuc tnh ca bng quan st(tip)

Nationality

Italian French German

Single Married

Rule 1: If (Nationality IS Italian) then (Conclusion IS B)

Rule 2: If (Nationality IS French) then (Conclusion IS A)
Rule 3: If (Nationality IS German) AND (Family IS Single)
then (Conclusion IS A)

Rule 4: If (Nationality IS German) AND (Family IS Married)

then (Conclusion IS B) Slide 109
5.3.2 Thut gii Hc theo bt nh

Stt Age Competition Type Profit

1 Old No Software Down
2 Midle Yes Software Down
3 Midle No Hardware Up
4 Old No Hardware Down
5 New No Hardware Up
6 New No Software Up
7 Midle No Software Up
8 New Yes Software Up
9 Midle Yes Hardware Down
10 Old Yes Hardware Down
Slide 110
Hc theo bt nh(tip)

bt nh ca X: k
E ( X ) = - p i log 2 p i
i =1

Tnh Entropy cho mi thuc tnh v chn thuc tnh c Entropy
nh nht.
k
E ( C / A ) = - p ( c i , a i ) log 2 p ( c i , a i )
i =1

4 4 2 2
E ( C / Competitio n = No ) = - log 2 - log 2 = 0 . 918
6 6 6 6
1 1 3 3
E ( C / Competitio n = Yes ) = - log 2 - log 2 = 0 . 811
4 4 4 4
E ( C / Competitio n ) = 0 . 6 * 0 .918 + 0 .4 * 0 . 811 = 0 .8752
Slide 111
Hc theo bt nh(tip)

Tng t:
E(C/Age) = 0.4 STT Competition Type Profit
E(C/Type) = 1 1 Yes Software Down

Age cho nhiu thng tin
nht 2 No Hardware Up
3 No Software Up
4 Yes Hardware Down
Age

Old Milde New

Down Competition Up
No Yes
Up Down Slide 112
Hc theo bt nh(tip)

Age

Old Milde New

Down Competition Up
No Yes
Up Down
Rule 1: If (Age IS Old) then (Profit IS Down)
Rule 2: If (Age IS New) then (Profit IS Up)
Rule 3: If (Age IS Midle) And (Competition IS No)
then (Profit IS Up)
Rule 4: If (Age IS Midle) And (Competition IS Yes)
then (Profit IS Down) Slide 113