Introduction to Deep Learning in Python
Learn how neural networks make and improve predictions, then build, validate and tune regression and classification models in Python with Keras.
This course shows how neural networks generate predictions from data and how to train them in Python. You code the forward pass yourself, adjust weights with gradient descent and backpropagation, then use Keras to specify, compile and fit regression and classification models. Later you validate them and experiment with wider and deeper networks, ending with a digit recognition model.
What You Will Achieve
- Code forward propagation through a multi-layer network that uses the rectified linear activation function
- Calculate model errors and slopes, then improve weights with gradient descent and backpropagation
- Specify, compile and fit Keras models for regression or classification and use them to make predictions
- Check a model against validation data, stop training early and test whether a wider or deeper structure works better
Program Overview
Neural Network Basics and Forward Propagation
Compares neural networks with classical regression and shows how values move through a network to produce predictions.
- Activation functions and ReLU
- Applying a network to many rows of data
- Deeper and multi-layer networks
- Learned levels of representation
Optimization with Backward Propagation
Measures how weight changes affect accuracy and uses that measure to improve a network’s predictions.
- Calculating model errors
- Gradient descent and slopes
- Making multiple weight updates
- How forward and backward propagation relate
Building and Fine-Tuning Keras Models
Applies the Specify-Compile-Fit workflow in Keras, then tunes models through validation and changes to their structure.
- Regression and classification models
- Optimization parameters and early stopping
- Model capacity, wider networks and added layers
- A digit recognition model on images
Certificate Awarded (Paid)
Job Outlook
Building and training neural networks in Python supports machine learning work on image recognition, natural language processing and robotics, as well as personal deep learning projects.
Good for: Python learners who want to understand how a neural network computes and learns by coding it by hand before building models in Keras.
Keep in mind: The lessons are written for the Keras 2.0 library and were last updated in 2022, so check the current Keras documentation when you apply the code.
Bottom line: Coding forward and backward propagation before turning to Keras gives a clear view of what the library automates for you.
Frequently Asked Questions
What role does Keras play in the course?
Keras is the Python library used in the second half to build deep learning models. You follow its Specify-Compile-Fit workflow for regression and classification, then validate the models, apply early stopping and try wider and deeper structures.
Why are forward and backward propagation taught before Keras?
The first two chapters build an understanding of how a network produces predictions and how backpropagation adjusts its weights. That foundation supports the second half, where Keras handles these steps while you build and tune models.
Does the course include work with image data?
Yes. The final chapter steps up to images, and after experimenting with wider networks and added layers you build your own digit recognition model.


There are no reviews yet.