TensorFlow and Keras

Keras Sequential API

Keras provides user-friendly neural network building.

Model Creation

model = Sequential([Dense(64, activation='relu'), Dense(10, activation='softmax')]).

Dense layers are fully connected. Activation specified in layer.

Training

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy']).

model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.2).

Evaluation

model.evaluate(X_test, y_test). predict(X_new) generates predictions.

model.summary() shows architecture.

Key Takeaways

1. Keras Sequential API builds simple models
2. compile/fit/train is standard workflow
3. Dense layers are fully connected