Football Predictions

  

Total views

2,559,625

Football Predictions (FPP) - The unique global solution that has the same test accuracy as training accuracy in machine learning with Auto AI. This ensures generalization for any model as well as the ability to reliably predict every event

See Football Predictions in Open Invention Network

Wed Aug 21 02:44:49 UTC 2019

Football Predictions 4.0.2 released!

Added the --test-fixed option:

This version also shows more clearly the size variation of the population

Sat Aug 17 08:12:00 UTC 2019

Football Predictions 4.0.1 released!

This version improves the contraction mapping algorithm.

Thu Aug 8 14:06:17 UTC 2019

Football Predictions 4.0 released!

Changes in this version:

1) Add Auto AI feature running in multithreaded mode
You can run up to 45 jobs simultaneously to optimize hyperparameters in a wide range.

2) Add embedding technique to handle sparse, high scale data like text classification.

3) Add optimization options

4) Add display of test accuracy and training accuracy, add display of variable size population

Below is an example of training the solution Sail with 5 jobs running simultaneously, optimizing with 32 hidden units and deep training. Test accuracy reaches 99%

./fpp -ld --units-min=5 --units-max=5 -j5

Fri Jul 5 04:34:27 UTC 2019

Football Predictions 3.0 released!

Changes in this version:
1) Use mini-batch method for big sample set
2) Extract test set from big sample set
For big sample sets, it is necessary to extract a test set. However when the sample set is small, the id3 method is still used. That is because for a small sample set, the nature of the problem may be changed when missing a few samples.
3) Add -r option to read input directly from command line parameter
This option is useful when running the program on a notebook. fpp-3.0 is used as an option in the Sail solution

4) Optimize hyperparameters using genetic algorithms
Network configuration is automatically established according to the principle of convergence to fixed point of genetic algorithm (Banach fixed point)
5) Add --id3 option to run fpp as an id3 program

6) Fields in the data file fppdata.txt are separated by commas (like .csv files) instead of spaces as before
This allows a field value to contain spaces. For example "General cargo".

7) Accept field values as numbers instead of only names as before
The input widgets for the numeric fields automatically change to input box instead of radio list

Sat Jun 23 08:02:51 UTC 2018

Football Predictions 2.0 released!
This is a great version. The program uses ID3 algorithm to generate some checking samples. Many neural network populations are processed with different number of hidden nodes, and through a number of different weight initialization. The populations evolve to their fixed point, when all neural networks have converged, the errors on the checking sample set is used to select population whose smallest error.

Mon Jun 18 11:35:39 UTC 2018

Football Predictions - id3 released!
The id3 program is intended to create checking set for Football Predictions 2.0. However, it can also run independently.

Sun May 6 06:09:17 UTC 2018

Football Predictions 1.1 released!
Increased the generalization.

Sat Sep 3 03:05:10 UTC 2016
Football Predictions 1.0 released!
This version uses genetic algorithms to optimize neural network
+--------------------+
Thu Jun 4 10:59:56 UTC 2015
Football Predictions Program - fpp 0.9.7.1
Extended to 25 factors and 10 output values
+--------------------+
Sun May 3 19:53:30 UTC 2015
Football Predictions Program - fpp 0.9.7
(Using the Neural Network without learning again)
Once learned, the program will save the results to a file named weights.txt. Next time, it will run the network instantly without having to learn again. To force the program to learn, please run the following command
./fpp --learn
+--------------------+
Thu Apr 30 17:17:40 UTC 2015
Football Predictions Program - fpp 0.9.6
(MODERN NEURAL NETWORK - Powerful automated learning machine)
This version focuses on Neural Network technology. It maintains three processing layers and uses logarithmic derivatives to supervise learning.
+--------------------+
Wed Apr 8 13:30:06 UTC 2015
Football Predictions Program - fpp 0.9.5 (Version of Optimized degree of learning)
This version optimizes the learning degree. Each problem has its individual learning degree in order to meet the requirements of multipurpose..
+--------------------+
Fri Apr 3 21:52:00 UTC 2015
Football Predictions Program - fpp 0.9.4 (universal version)
This version is programmable. You can define factors and their properties, put sample data to solve your own problem. Everything need to do is to compose fppdata.txt file.
Fpp can be a Football predictions program, a Weather forecast program... all in one.
+--------------------+
Sun Mar 29 21:19:58 UTC 2015
Football Predictions Program - fpp 0.9.3
Supported user data file. You should make a file named fppdata.txt to store samples.
Limitation: custom data but five factors only with the fixed format
+--------------------+
Fri Mar 27 08:57:02 UTC 2015
Football Predictions Program - fpp 0.9.2
Accepted that several samples won’t match the calculating to obtain the logic in general.
The program “understands” the events.
+--------------------+
Tue Mar 24 23:52:56 UTC 2015
Football Predictions Program - fpp 0.9.1
This version calculates 5 factors with data samples of 16 matches
+--------------------+
Mon Mar 23 08:13:22 UTC 2015
Football Predictions Program - fpp 0.9.0
Usage: ./fpp (you should make it executable first: chmod u+x fpp)
Once tried to run fpp and like it please do not hesitate to contact me
+--------------------+
Thanks for your consideration!
Pham Thanh Tuyen <phamtyn@gmail.com>
Tel: +84 913509520