AIML

Scikit Learn: a general machine learning library (built on top of NumPy) (not a neural network)
PyTorch: deep learning framework (neural network)
1. Suitable for:
  1. Flexibility
  2. Short training duration
  3. Debugging capabilities
Keras: higher-level deep learning framework (making code simpler and more concise than PyTorch and Tensorflow).
1. High-level, run on top of Tensorflow. Interface rather than a standalone framework.
2. Tensorflow supports Keras with tf.keras library
3. Suitable for:
  1. Rapid prototyping
  2. Small dataset
  3. Multiple back-end support
Tensorflow: open-source software library for dataflow programming (especially with data in form of tensor)
1. Compatible with wide range of hardware and devices
2. Formats
  1. TensorFlow Lite
  2. TensorFlow.js
  3. TensorFlow Extended
3. Suitable for:
  1. Large dataset
  2. High performance
  3. Functionality
  4. Object detection
Ways to adding labels to images
1. From csv file with file path and label for each file
2. From folder’s names
3. Manually correct on the console itself
Images can be uploaded as a zip file.
1. Additional images can be uploaded to “unlabeled” folder or with
Images can be exported to an csv file with labels, data types (training, test)

Input type
1. Scalar, Vector, Matrix, Tensor (multi-dimensional matrix)
NNLayers
1. Input layer (layer 0)
2. Hidden layers
3. Output layer (e.g. output values)
Neuron
1. Input vector (X)
2. Output (Y)
  1. f (W(transpose) * X) = Y
3. Weight vector (W)
Architectures
1. Feed Forward NN
2. Recurrent NN
3. Convolution NN
4. GAN: generative adversarial network

1. Delete a model (in case the GUI bugs out)
gcloud ai models list
gcloud ai models delete $MODEL_ID