Introduction to TensorFlow and Keras
After understanding neural networks, perceptrons, activation functions,
and backpropagation, it’s time to explore how deep learning is built
in the real world. The most popular and powerful deep learning framework
today is TensorFlow, and its high-level API
Keras makes model building extremely easy.
In this chapter, you will learn:
- What is TensorFlow?
- What are Tensors?
- Why Keras makes deep learning easier
- How to build your first neural network
- Training, evaluation, and predictions
- Real-life examples using TensorFlow/Keras
⭐ What Is TensorFlow?
TensorFlow is an open-source deep learning framework created by Google.
Most of Google’s AI systems are powered by TensorFlow, including:
- Google Photos
- Google Translate
- YouTube recommendations
- Google Assistant
- Google Lens
- Self-driving car technology
TensorFlow helps developers build neural networks, train them,
and deploy them in production at massive scale.
📌 Why It’s Called “TensorFlow”
A tensor is a multi-dimensional array (like a matrix).
The word flow means these tensors move through a computation graph.
Tensor + Flow = TensorFlow.
Example:
- Scalar → 0D tensor
- Vector → 1D tensor
- Matrix → 2D tensor
- Images → 3D/4D tensor
📌 Real-Life Example of Tensors
When you upload an image to Google Photos:
- Height × Width × Color channels (RGB)
This becomes a tensor, which deep learning models use for prediction.
⭐ What Is Keras?
Keras is a simple, user-friendly API built on top of TensorFlow.
Instead of writing complex code, developers can build models in a
few lines using Keras.
Keras = Easy Deep Learning
It allows you to:
- Create layers easily
- Build neural networks quickly
- Train models with simple commands
- Evaluate and test models efficiently
📌 Why Developers Prefer Keras
- Simplest API for beginners
- Clean and readable code
- Runs on top of TensorFlow (very powerful)
- Great for research and production
⭐ Your First Program: Importing TensorFlow
import tensorflow as tf
print(tf.__version__)
This prints the installed TensorFlow version.
📌 Creating and Working with Tensors
Creating a tensor:
import tensorflow as tf
a = tf.constant([1, 2, 3])
print(a)
Tensors can store images, text embeddings, audio waveforms, etc.
⭐ Building Your First Neural Network (Keras)
Let’s create a simple neural network for digit classification.
from tensorflow import keras
from keras import layers
model = keras.Sequential([
layers.Dense(64, activation='relu'),
layers.Dense(10, activation='softmax')
])
This model has:
- 1 hidden layer with 64 neurons
- Output layer with 10 classes (0–9)
- ReLU and Softmax activations
