Skip to content
Different levels of abstraction in convolutional neural network implementations with TensorFlow
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


Differently abstract implementations of an exemplary convolutional neural network with TensorFlow.

  • Convolutional layer from scratch: Convolutional neural network implementation with convolutional and pooling layers built from scratch with core TensorFlow.

  • Low-level TensorFlow: Convolutional neural network using tf.nn.conv2d and tf.nn.avg_pool with explicit definitions of weights, biases, and placeholders.

  • Mid-level TensorFlow: Convolutional neural network using tf.keras.layers managing weights and biases for us, whereas placeholders and the session are still explicit.

  • High-level TensorFlow: Convolutional neural network using tf.keras.model.Sequential (everything is managed).

You can’t perform that action at this time.