Hello everyone, we hope you will contribute for C++ Online Course that was initiated by daynhauhoc.com and free for everyone who need learn about C++. I am contributing to OOP C++ Lessons. That is community programming. :)
You can fork the project, contribute to it by yourself and pull request to submit your lesson :)
Thank you all!
https://github.com/daynhauhoc/cppoop
Ảnh
Chủ Nhật, 15 tháng 5, 2016
Thứ Hai, 2 tháng 5, 2016
My Autocomplete
Autocomplete on PC. Its GUI work very well but I disabled it :) It can "learn" user behavior to predict exacly the words that user want to type. I crawled a lot of paragraphs from wikipedia, pushed them into the program to train it automatically. I also referenced some solution from Unikey. (Unikey is an open source), thanks for Unikey dev team :)
The first time I implemented an AI :)
This is the first time I created a bot what can automatically play some games. Fortunately, my Bot worked better a lot than I would. Mancala (called "Ô Ăn Quan") is the game I was choosen to implement an AI.
At first, I just wanted to make an simple solution for AI by minimax, alpha-beta prunning or ID3. Everything was OK and I collected a lot of data in every rounds between player and bot. So I think it can be used for training an ANN (Artificial Neural Network). I have been tried push all data that I collected into Weka that use to analysis data. My data involved 12 input, so I made an network with 12 input nodes and 1 output node. The package I used to create a neural network is MutilLayer Perceptron, that neural network is make by one hidden layer (include 6 sigmoid function nodes) and last layer (include a linear function node). As a result, I had all weight in neural network, and put them all on my BOT. After 100 tests, it not just worked very well but also beat all solution that I implemented. Proud of MY FIRST BOT :)).
This network is just a small neural network, a prediction system is common work with a neural network of over 500 nodes (handwriting recoginiton system also commonly works with such level).
Btw, it worked very well and made excited for my mentor.
At first, I just wanted to make an simple solution for AI by minimax, alpha-beta prunning or ID3. Everything was OK and I collected a lot of data in every rounds between player and bot. So I think it can be used for training an ANN (Artificial Neural Network). I have been tried push all data that I collected into Weka that use to analysis data. My data involved 12 input, so I made an network with 12 input nodes and 1 output node. The package I used to create a neural network is MutilLayer Perceptron, that neural network is make by one hidden layer (include 6 sigmoid function nodes) and last layer (include a linear function node). As a result, I had all weight in neural network, and put them all on my BOT. After 100 tests, it not just worked very well but also beat all solution that I implemented. Proud of MY FIRST BOT :)).
This network is just a small neural network, a prediction system is common work with a neural network of over 500 nodes (handwriting recoginiton system also commonly works with such level).
Btw, it worked very well and made excited for my mentor.
 |
| my Mancala game |
 |
| The neural network that I made for my bot |
Đăng ký:
Bài đăng (Atom)
