Menu
I'm working on a convolutional neural network and I've found various methods of building filters to convolve the image with. What are the advantages and disadvantages of each?
649
That's a lot of data, processing time, hyperparameter tuning, and model validation before you even start building the real model. An autoencoder learns to efficiently represent a manifold on which the training data lies.
Another drawback is you may eliminate the important information in the input data. The autoencoder algorithm requires an objective function for evaluating the accuracy of encoded/decoded input data. In most applications it’s not possible to represent the true objective, and some simple proxy like sum-of-squared-errors is used.
You have to train an autoencoder. That's a lot of data, processing time, hyperparameter tuning, and model validation before you even start building the real model. An autoencoder learns to efficiently represent a manifold on which the training data lies. If your training data is n Autoencoders can be great for feature extraction.
This requirement dictates the structure of the Auto-encoder as a bottleneck. The Auto-encoder first tries to encode the data using the initialized weights and biases. The Auto-encoder tries to reconstruct the original input from the encoded data to test the reliability of the encoding.
Using autoencoders may in some cases improve performance, yield biologically plausible filters, and more importantly, give you a model based on your data instead of predefined filters. Autoencoders will give you filters that may fit your data better, in general. The only downside is of course the additional computation time. However, in many cases it doesn't hinder you from learning online.
For more evaluations you might have a look at: http://www.idsia.ch/~masci/papers/2011_icann.pdf
93
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
