## Neural Network Structure

From the previous chapter, we've seen how to create individual layers. It's important to understand that these layers become effective only when they are **integrated into a cohesive model**.

### Sequential Model

In Keras, a straightforward way to construct a model is by using the `Sequential` model. It's essentially a **linear stack of layers**.

> **Note**
>
> You can use `model.summary()` to view the model's structure.

### Method 1

Initiate a `Sequential` model by passing a list of layers to the **constructor**.

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Activation, Input

# Create a model with layers
model = Sequential([
    Input(shape=(2,)),
    Dense(units=3),
    Activation('relu'),
])

model.summary()

### Method 2

Alternatively, start with an empty `Sequential` model and **add layers post-creation**.

> **Note**
>
> This approach is generally more flexible, allowing for **dynamic model assembly**. Due to its adaptability, it's often the preferred method.

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Activation, Input

# Create a model
model = Sequential()

# Add layers
model.add(Input(shape=(2,)))
model.add(Dense(units=3))
model.add(Activation('relu'))

model.summary()

Dive into the world of neural network development with an immersive program that blends foundational knowledge with advanced techniques. The course emphasizes hands-on learning, providing learners with the opportunity to apply these techniques. Additionally, you will face numerous data preprocessing tasks and real-world examples of neural network training, so stay prepared and motivated, as it will present a significant challenge for you.

This segment introduces the fundamentals of Keras through hands-on experiences with the Spotify dataset, covering everything from layer configuration and model creation to compilation, data preprocessing, and training.

The regularization module, utilizing the Mushrooms dataset, addresses the critical issues of overfitting and underfitting, explaining the concept of regularization and its importance in model training.

Exploring cutting-edge methodologies, this section delves into the Diamonds dataset to uncover the intricacies of optimizers, learning rate adjustments, TensorFlow Datasets, data generators, and advanced modeling techniques like non-sequential models, transfer learning, and multitask learning in a comprehensive overview.

Neural Networks with TensorFlow

Model Creation

Neural Network Structure

Sequential Model

Method 1

Method 2

Both methods yield the same model.