A pattern recognition & machine learning package for Python. It provides:
An autoencoder is a type of neural networks. Typically, it constists of same-sized input & output layers and some hidden layers with less number of units and is trained to be able to restore the input. As the result, the hidden layer's activation pattern, which has less dimensions than the input, can be interpreted to hold "essential" information about the input patter in a sense. Therefore, it is used for feature extraction or dimensionality reduction.
Here's an example using the MNIST handwritten digits dataset. See tests/test_ae.py
for the details.
Input | Output |
---|---|
Here're some examples for clustering algorithms. See tests/test_cluster2.py
for the details.
K-means | Competitive Learning | EM Algorithm |
---|---|---|