Tim Blog

Himmel oder Hölle

tensorflow functions

tf.cast tf.argmax tf.argmin tf.equal tf.cast() Casts a tensor to a new type tf.cast( x, dtype, name=None ) tf.argmax() Returns the index with the largest value a...

Posted by neverset on April 26, 2020

tensorflow Autoencoder

Autoencoder(AE) is an unsupervised learning method. it can be used to compress the original input information, so that learning is eased with reduced data size. At the same time the new encoded inp...

Posted by neverset on April 18, 2020

tensorflow Siamese Network

siamese network is a network that weights can be shared between layers usage this network can be used to decide the similarities between two inputs depending on the input the network structure ...

Posted by neverset on April 18, 2020

tensorflow Filter Response Normalization

To avoid that normalization dependes on batch size will lead to bad results when batch size on test and training is different, filter Response Normalization is introduced structure Filter Response...

Posted by neverset on April 18, 2020

tensorflow NIN

Network in Network is a micro network to replace the pure conv laver to extract more complex and non-linear features structure the difference between traditionale conv layer and mlpconv layer is ...

Posted by neverset on April 18, 2020

tensorflow ResNet

ResNet consists of multi shallow network blocks (residual blocks) the purpose of ResNet is to keep model from degradation if NN depth increases advantage: deeper networks solve model degradati...

Posted by neverset on April 13, 2020

tensorflow Global Average Pooling

Global Average Pooling (GAP) is an replacement of FC layer, which has two benefits more native to categorize feature map less parameter to optimize to avoid overfitting principle traditio...

Posted by neverset on April 12, 2020

tensorflow DenseNet

Densely Connected Convolutional Networks (DenseNets) Each layer has direct access to the gradients from the loss function and the original input signal, leading to an implicit deep supervision ...

Posted by neverset on April 12, 2020

tensorflow summary usage

summary method can save the training process and relevant parameters, which can be later visualized in tensorboard besides the loss and accuracy data in also be saved in np format for visualization...

Posted by neverset on April 12, 2020

tensorflow multi thread computing

tensorflow offers two classes for multi threads computing: tf.Coordinator it can stop all subthreads in the queue and report exceptions to the main thread tf.QueueRunner it starts multi thread...

Posted by neverset on April 12, 2020