/10DaysForCNN

Learning CNN in 10 days with students

Primary LanguageJupyter Notebook

10 Days For CNN

Keras Keras

Rules

  • Everyday 1 neuro network no matter what.

Links

DataSources

tf.keras.datasets.

boston_housing module: Public API for tf.keras.datasets.boston_housing namespace.

cifar10 module: Public API for tf.keras.datasets.cifar10 namespace.

cifar100 module: Public API for tf.keras.datasets.cifar100 namespace.

fashion_mnist module: Public API for tf.keras.datasets.fashion_mnist namespace.

imdb module: Public API for tf.keras.datasets.imdb namespace.

mnist module: Public API for tf.keras.datasets.mnist namespace.

reuters module: Public API for tf.keras.datasets.reuters namespace.

  • Day 0 Practical Use
  • Day 1 Simple Use cases

Day 0

todo

  • TensorFlow official web page
  • Jupyter notebooks
  • Course I recomended from rutracker
  • Keggle

Day 1

Form follow the shape

  • Metrics:
    • Time
    • Countity
    • Speed
    • CNN per week
    • Easiest way possible

Noted

traing_images.shape

(6000,28,28)

Black and white plot

plt.imshow(tri[i],cmap=plt.cm.binary)

Layers

Layer Types

  • Flatten - Matrix into array
  • Dense - comapring data layer
  • Softmax - Get total probability (typically last label)

Optimizer

model.compile(optimizer='adam',