-
OS: Ubuntu14.04
-
Language: Python3.5.2
I recommend using pyenv and venv.
$ pyenv install 3.5.2
$ pyenv local 3.5.2
$ pyvenv venv
$ . venv/bin/activate
(venv) $ pip install --upgrade pip
(venv) $ pip3 install -r requirements.txt
When you'd like to deactivate the environment,
(venv) $ deactivate
Then, install TensorFlow.
# Ubuntu/Linux 64-bit, CPU only, Python 3.5
(venv) $ export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.11.0-cp35-cp35m-linux_x86_64.whl
# Ubuntu/Linux 64-bit, GPU enabled, Python 3.5(Requires CUDA toolkit 8.0 and CuDNN v5.)
(venv) $ export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow-0.11.0-cp35-cp35m-linux_x86_64.whl
# Mac OS X, CPU only, Python 3.4 or 3.5:
(venv) $ export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-0.11.0-py3-none-any.whl
(venv) $ pip3 install --upgrade $TF_BINARY_URL
If you have trouble, click here for detail.
Move to source directory.
First, run this script to create a dataset.
(venv) $ python dataset.py
Make sure that there is height.pkl in the current directory.
This dataset has 1000 samples and each sample has 1-dimensional feature and its class label.
This is the simulation of human height.
Then, train VAE.
(venv) $ python vae.py
Check the loss-transition using tensorboard.
(venv) $ tensorboard --logdir=<absolute path to current directory>/experiment
Open your browser, and get access to localhost:6006.
You can show the loss-transition and the computational graph.
Extract encoder model and make its outputs features of samples.
The features represent distributions that generate the original samples.
Now, classify the samples using SVM.
(venv) $ python classify.py