Proof of Concept purpose is to prove economic benefits of final ML models merging. Neural network was chosen for several tests with text, audio and image datasets. The required time for training and merging the neural networks was compared. Neural networks merging allows to get a network with the necessary features 45-200 times faster than training a new neural network.
- numpy
- scypi
- tensorflow
- keras
- librosa
- cifar10
- cifar100
- audio 10 classes
- text 20 classes