Genetic Algorithm

Sarah_Radzi

adakah sesapa di sini kenal dengan Genetic Algorthm utk estimate data ?

i baru je dengar benda ni ...
ade sesapa tahu ?

Remy_3D

http://www.rennard.org/alife/english/gavintrgb.html#rTit01

Satria-Naga

Originally posted by Sarah_Radzi at 20-11-2003 12:58 AM:
adakah sesapa di sini kenal dengan Genetic Algorthm utk estimate data ?

i baru je dengar benda ni ...
ade sesapa tahu ?

maksud u analisis kod DNA ke?

Sarah_Radzi

computer algorithm yang imitate cara hidup dna  ..

Satria-Naga

Originally posted by Sarah_Radzi at 29-11-2003 07:49 PM:
computer algorithm yang imitate cara hidup dna  ..

bab computer.. i tak tau...

atira

GA ni salah satu method of AI kan...
kalau sarah ni cakap GA untuk estimate data...
so.. aku rasa maksud dia camni... dalam satu area contohnye la business kan... untuk forecast kita punya business m/c or program tu boleh estimate kan data untuk kita.. so program tu kene paham keadaan business kita, keadaan pasaran etc etc etc..
jadi method kita nak ajar program/ machine tu, tu lah guna GA..
harap2 aku tak kasi ajaran sesat.. he he

Sarah_Radzi

u tak kasi ajaran sesat atira
betul lah tu
cuma skop u n i berbeza ..

mareenie

Genetic Algorithm (GA) melibatkan penggunaan konsep penggabungan kromosom yang diaplikasi dlm bnyk bidang sains.salah satu yang bnyk digunakan ialah dlm teknik penjadualan pengeluaran(shop floor scheduling).

nak tulis lagi..tp busy aa..nnt ek..

So, kena idetify conditions yg dominant and resesive.

Dalam 4 keuntungan mungkin ada satu kerugian? (klu biznes la).

mareenie

Originally posted by Daffodill at 14-1-2004 07:32 PM:
So, kena idetify conditions yg dominant and resesive.

Dalam 4 keuntungan mungkin ada satu kerugian? (klu biznes la).

GA adlh kaedah pencarian yg digunakan sbg alat penyelesaian bg ,mslh pengoptimuman..eg: pengoptimuman gabungan. Konsepnya sama dgn konsep percantuman kromosom...katalah dlm keturunan kita 1000 jenis kromosom yg masing2 membawa bnyk jenis gen (yg mewakili warna mata,jenis rambut,ketinggian,etc)....tiap2 jenis kromosom tu pulak masing-masing  membawa gen ciri dominan atau gen berciri resesif ... tiap2 kromosom ni mesti combine each other utk hasilkan 1 combination yg terbaik (cthnya yg anda bw dlm diri anda skrg)....
contoh:
kromosom dgn gen rambut kerinting(dominan)+kromosom dgn gen warna kulit gelap(resesif)+kromosom dgn warna mata biru(resesif).
combination yg lain yg boleh jd mungkin:
kromosom dgn rambut kerinting(dominan)+kromosom dgn gen warna kulit gelap(resesif)+kromosom dgn warna mata hitam(dominan).
and so on..

jadi, kebarangkalian jumlah combination yg terhasil utk menghasilkan satu kromosom yng mengandungi gen-gen terbaik= 1000! (1000 faktorial)....

...dlm 1 keuntungan mesti ada kerugian...hmmmmmmm....katalah u are a pen manufacturer and u nak produce satu model pen yg baru ...dan satu pen tu ada part2 yang kena assembly kan...ok..let say u want to decide either u want to buy or make a part on your own. here u have choice...buy and make (utk gantikan istilah dominan & resesif - mana yg baik & mana yg tak  baik i assume anda sendiri sudah tahu ...)
dalam 1 pen tu, ada 8 part...
so kebarangkalian pilihan yg ada nnt ada lebih kurang 8!=40320...
jadi pilihan yg anda pilih mungkin.. beli semua part pd pen tu, then anda assemble. tak pun...semua part pd pen tu ada buat sendiri, yg anda beli cuma rubber saja...there are a lot of choices!

ini contoh mudah je lah yg sy mampu terangkan.....kalau nak citer yg lg complex takut org tak paham pulak ye...

[ Last edited by mareenie on 16-1-2004 at 05:49 AM ]

hippo_02

hemm..bagaimana pulak kalau nak applykan ga ni dalam domain seperti clustering chemical compound.nak optimize dari segii kepelbagaian (diversity) sesuatu cluster itu dengan menggunakan ga.does anybody know how to create the fitness function base on this domain?:stp:

mareenie

hippo_02

your question is quite tough for me...got to ask my friends about it..

adlismel

GA adalah sebahagian daripda subjek Artificial Intelligent......

dulu masa saya belajar, pernah dengar juga Genetic Algorithm ni , kalau tak salah saya ia adalah satu kaedah nak cari penyelesaian terbaik bagi sesuatu masalah dengan menggabungkan beberapa kaedah penyelesaian yang lain, my fren ni dia guna MATLAB utk buat Genetic Algorithm programming , tak salah saya Genetic Algorithm ni macam Linear Programming gak , i dunno elok rujuk la , dari apa yg saya tahu Genetic Algorithm ni menggabungkan kaedah numerical method (iteration by iteration) dan biology (sumthing to do with DNA lor) , Monte Carlo method pun dalam group yang sama , emm ini banyak kaitan dgn maths sebennarny , cuba refer balik la kat mana2 website , penah tengah cerita a beautiful mind? (russel crowas Prof Nash)

mbhcsf

salam ..
it sounded like DNA computing

Sarah_Radzi

Originally posted by ^clump^ at Tue 04-Jan-2005 14:46:
dulu masa saya belajar, pernah dengar juga Genetic Algorithm ni , kalau tak salah saya ia adalah satu kaedah nak cari penyelesaian terbaik bagi sesuatu masalah dengan menggabungkan beberapa kaedah  ...

ha'ah .. clump betul ...
itu ah .. i figured it out walaupun tak sempat lagi nak mendalami ..

Originally posted by mbhcsf at 5-1-2005 11:20 AM:
salam ..
it sounded like DNA computing

but its not.

Originally posted by ^clump^ at 4-1-2005 08:46 PM:
dari apa yg saya tahu Genetic Algorithm ni menggabungkan kaedah numerical method (iteration by iteration) dan biology (sumthing to do with DNA lor) , Monte Carlo method pun dalam group yang sama , emm ini banyak kaitan dgn maths sebennarnya

GA inspired by theori Darwin(rujuk posting mareenie, he/she was describing natural selection process) wlwpun begitu, you don't need to know anything about biology atau dna to use GA, other than knowing some of biology terms(phew!).

Lets start by knowing some MUST-KNOW terms:
Nature/In biology             Computer
Individual                        Solution to a problem
Population                      Set of solutions
Fitness                           Quality of a solution
Chromosome                   Encoding for a solution
Gene                             Part of the encoding of the solution
Crossover & Mutation       Search operators
Natural Selection             Reuse of good solutions

Lets now leap on by introducing a generic GA:
1. Initialize and evaluate a population.
2. While (termination condition not met) do
        a. Select sub-population based on fitness
        b. Produce offspring of the population using crossover operator
        c. Mutate offspring stochastically
        d. Select survivors based on fitness

We'll discuss every step of the algorithm more next time, but for now, lets note that the algorithm tries adopt the concept of real life phenomena in solving optimisation problems. Example of problems ranged from the simplest such as finding the value of x which maximises the function f(x)=x^2 on the integer range of [0...31], to the complex such as coming up with university's exam timetable to avoid clashes(scheduling problems).

Sarah, you've been digging this subject for more than a year now. How's the research going? What kind of data were u estimating?

To be continued...

lupa pulak, Monte Carlo lebih kurang sama dgn GA in the sense that both applied to a complex problem that we do not have to know the complete model of the problem to actually start solving it. BUT, Monte Carlo and GA are both completely different domains. You have to actually convert/encode a solution to bits in order to use GA and apply the genetic operators(crossover and mutation) and use fitness functions to come up with solutions. Monte Carlo is a different way of finding solution. It worths a topic of its own under machine learning. ff::cak:

Sarah_Radzi

it turn out that i'll take too much time understanding and applying the algorithm to my problems ... i had very strict time constraint . so i  abandon GA at that time and for the time being ..

ps: data estimation concerning digital communication/signal process