Move from TensorFlow 2 fundamentals to working neural-network models across nine Jupyter notebooks. You will begin with tensors, eager execution, and automatic differentiation, then implement regression and classification models before progressing to convolutional networks, transfer learning, and model persistence.
The course deliberately compares several levels of the TensorFlow API. You will write low-level calculations with variables and GradientTape, construct models with tf.keras Sequential and Functional patterns, subclass models, and complete focused implementation tasks using real and synthetic data.
What You Will Learn
- Create and manipulate TensorFlow tensors and variables in eager execution
- Compute derivatives and track gradients with
tf.GradientTape - Implement linear regression with low-level operations, layers, and Keras models
- Fit polynomial regression by completing a low-level TensorFlow implementation
- Build shallow classifiers with custom code, Sequential, Functional, and subclassed models
- Prepare and classify handwritten-digit and car-safety data
- Construct VGG-style and LeNet-5 networks for Fashion-MNIST and adapt a pretrained VGG16 base
- Save and restore weights, complete models, and low-level checkpoints
Who This Course Is For
This course is for learners who know Python and introductory machine learning and want a code-first tour of TensorFlow 2 from tensors to convolutional networks. The notebooks are labeled beginner but cover substantial ground quickly, so they suit learners who already understand arrays, regression, classification, and the basic purpose of neural-network layers. Deployment and production serving are not covered.
Prerequisites: Basic Python, NumPy-style array operations, and introductory knowledge of derivatives, regression, classification, and neural networks are recommended. Familiarity with Jupyter Notebook is helpful.
Learning environment: All nine activities run in a provided Ubuntu 22.04 Jupyter environment. Prepared notebooks contain explanations, runnable examples, data-loading code, and implementation sections for you to complete.
Frequently Asked Questions
Is this only an introduction to TensorFlow syntax?
No. It starts with syntax and automatic differentiation, then moves through regression, shallow classification, VGG-style and LeNet-5 convolutional models, transfer learning, and model saving.
Which data sets are used?
The notebooks use generated regression data, scikit-learn’s handwritten digits, the Car Evaluation data set, and Fashion-MNIST. These support regression, tabular classification, and image classification exercises.
Does every example use the latest TensorFlow 2 API style?
No. The main approach is TensorFlow 2—eager execution, GradientTape, and tf.keras—but some notebook cells retain early TensorFlow 2-era names such as tf.train.AdamOptimizer. Check current TensorFlow documentation when adapting the examples to a new environment.
Is there one final project?
No. Instead, the course includes focused code-completion tasks for polynomial regression, car-safety classification, and a LeNet-5-style network, alongside the guided notebooks.
