COLA - Competitive layers for deep learning

COLA (COmpetitive LAyers) is a Python package providing the implementation of gradient-based competitive layers which can be used on top of deep learning models for unsupervised tasks.

Theory

Theoretical foundations can be found in our paper.

If you find COLA useful in your research, please consider citing the following paper:

@misc{barbiero2020topological,
    title={Topological Gradient-based Competitive Learning},
    author={Pietro Barbiero and Gabriele Ciravegna and Vincenzo Randazzo and Giansalvo Cirrincione},
    year={2020},
    eprint={2008.09477},
    archivePrefix={arXiv},
    primaryClass={stat.ML}
}

Examples

Dual Competitive Layer (DCL)

Vanilla Competitive Layer (VCL)

Using COLA

from cola import DualModel, plot_confusion_matrix, scatterplot, scatterplot_dynamic

X, y = ... # load dataset

# load custom tensorflow layers
inputs = Input(shape=(d,), name='input')
...
outputs = ...

# instantiate the dual model
n = X.shape[0] # number of samples
k = ... # upper bound of the desired number of prototypes
model = DualModel(n_samples=n, k_prototypes=k, inputs=inputs, outputs=outputs, deep=False)
model.compile(optimizer=optimizer)
model.fit(X, y, epochs=epochs)

# plot prototype dynamics
plt.figure()
scatterplot_dynamic(X, model.prototypes_, y, valid=True)
plt.show()

# plot confusion matrix
# considering the prototypes estimated in the last epoch
plt.figure()
plot_confusion_matrix(x_pred, model.prototypes[-1], y)
plt.show()

# plot estimated topology
# considering the prototypes estimated in the last epoch
plt.figure()
scatterplot(x_pred, model.prototypes[-1], y, valid=True)
plt.show()

Authors

Pietro Barbiero

Licence

Copyright 2020 Pietro Barbiero.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at: http://www.apache.org/licenses/LICENSE-2.0.

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

See the License for the specific language governing permissions and limitations under the License.