What is Keras?

Keras is a high-level Deep Learning API written in Python that allows developers to quickly build, train, and deploy Neural Networks. It runs on top of TensorFlow and is now an integral part of TensorFlow 2.x.

Why is Keras Important?

• Provides a simple and clean API for Deep Learning.
• Built directly on TensorFlow — fast and scalable.
• Reduces development time significantly.
• Great for beginners and researchers.
• Used by Google, Netflix, Uber, and many big companies.

History of Keras

Why Use Keras?

Core Features of Keras

Keras Architecture

Keras follows a clean, layered architecture:

Types of Keras APIs

How to Install Keras

pip install tensorflow

Keras is included inside TensorFlow:

import tensorflow as tf
from tensorflow import keras
print("Keras version:", keras.__version__)

Building a Model — Sequential API

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
import numpy as np

# Sample data
X = np.array([[0,0],[0,1],[1,0],[1,1]])
y = np.array([0, 1, 1, 1])

# Sequential model
model = Sequential([
    Dense(8, activation="relu", input_shape=(2,)),
    Dense(1, activation="sigmoid")
])

# Compile model
model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])

# Train model
model.fit(X, y, epochs=200, verbose=0)

# Predict
print(model.predict(np.array([[1, 0]])))

Functional API Example

from tensorflow.keras.layers import Input, Dense
from tensorflow.keras.models import Model

inputs = Input(shape=(2,))
x = Dense(8, activation="relu")(inputs)
outputs = Dense(1, activation="sigmoid")(x)

model = Model(inputs=inputs, outputs=outputs)
model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])
model.summary()

Common Layers in Keras

Common Activation Functions

• ReLU — Default for hidden layers
• Sigmoid — Binary classification
• Softmax — Multi-class classification
• Tanh — Hidden layers
• Leaky ReLU — Avoids dead neurons

Model Compilation

Three key components used when compiling a model:

• Loss function — Measures error.
• Optimizer — Updates weights.
• Metric — Evaluates performance.

model.compile(
    optimizer="adam",
    loss="binary_crossentropy",
    metrics=["accuracy"]
)

Training a Keras Model

model.fit(X, y, epochs=100, batch_size=4, validation_split=0.2)

Evaluating & Predicting

# Evaluate
loss, accuracy = model.evaluate(X, y)
print("Accuracy:", accuracy)

# Predict
predictions = model.predict(X)
print(predictions)

Save & Load a Model

# Save
model.save("model.h5")

# Load
from tensorflow.keras.models import load_model
loaded_model = load_model("model.h5")

Real-Life Analogy

Keras vs TensorFlow

Real-World Applications

Advantages of Keras

• Easy to learn and use.
• Highly modular.
• Supports CPU, GPU, TPU.
• Excellent documentation.
• Backed by TensorFlow.

Disadvantages

Common Mistakes to Avoid

Best Practices

Importance of Keras

Golden Rule