A program where given a sequence of input-output pairs, tries to generate a linear function that best simulates the algorithm using a local greedy hill-climbing approach.
Current working dir: ./bin
Reading from "./bin/../data/data.txt
There are 3 inputs
There are 3 outputs
Building Convolutional Neural Net with 3 networks
Starting setup...Current working dir: ./bin
Reading from "./bin/../data/data.txt
Read file successfully
..Initializing network: 0... SUCCESS!
..Initializing network: 1... SUCCESS!
..Initializing network: 2... SUCCESS!
Network: 0 is built with:
-- 3 -- 3 -- 4 -- 1 neurons
Network: 1 is built with:
-- 3 -- 2 -- 2 -- 2 -- 1 neurons
Network: 2 is built with:
-- 3 -- 2 -- 2 -- 1 neurons
setup complete.
Welcome to Convolutional Neural Net v1.0.0
Type (Help) or (H) or (?) to see all options and input commands (Ignore parens)
Inputs:
(test) ===> input new data and compare
(save) ===> save current network data to output.txt
(cost) ===> print average network cost
(train) ===> begin training of neural net
(multi) ===> use multithreading computation
(single) ===> use single-core computation
(reset) ===> reset network and randomize all weights
(exit) ===> exit from program(1)Next Command:
>>
- OpenFrameworks v0.10.1+ openframeworks.cc/download/ (Or look into
OF_install.sh
thats provided) - Unzip and place in your home directory with name
OF/
or edit the OF_ROOT in the Makefile - Run the internal script inside
OF/scripts/linux
or similarly for the windows/MacOS installation
g++
with a terminal
For the graphical interface you will need OpenFrameworks and there is an automated script (OF_install.sh) that will download everything and setup the OF directory as required by the program. The script was written on a Ubuntu derivative, however to run on Arch/fedora/Debian there are similar install_dependencies scripts in the OF directory (refer to the OF_install.sh script). Then run install_graphical.sh
in the project directory, then enter bin
to run the graphical
executable.
Execute make_tty.sh
in the project directory to create the binary and enter bin
to run ConvNeuralNet
To use custom datasets simply modify/replace the data.txt file in data/
.
Things to note:
- The format of the dataset should be "x1 x2 ... xn ; y1 y2 ... ym " on every line (m dosent have to equal n)\
- Spaces are used to divide the individual data entries and the semicolon (;) separates the inputs against outputs.
- Make sure to use the same number of inputs for all data entries (as would a normal function mapping).
PRESS (H) (h) or (?) to see this from the terminal
- SPACE: Iterate through dataset and run into model.
- BACKSPACE: Compute total average cost, where cost function is the difference in squares
- RT_SHIFT: Begin training on dataset
- ENTER: Randomize weights on entire Neural Net\
- ESC: Exit application
- All the information for input commands should be presented upon program launch. Or simply type 'help', 'H', or '?'