Open Source Deep Learning Course from Universitat Autònoma de Barcelona: A Complete Learning Path from Perceptrons to Transformers

The DL2025-26 core course of the Artificial Intelligence degree at Universitat Autònoma de Barcelona (UAB) has been officially open-sourced on GitHub, providing a systematic deep learning knowledge system from perceptrons to Transformers for AI learners worldwide. The course integrates theory and practice, covering core modules such as neural network fundamentals, CNN, RNN/LSTM, and Transformer, with abundant programming assignments and resources. It is suitable for learners from zero to intermediate levels and is a rare opportunity to access university-level deep learning education for free.

In today's era of rapid AI development, deep learning is the core driving force, but learners often face the challenge of lacking a systematic learning path. As a top research university in Spain, UAB's AI degree program has received much attention. This open-sourcing of the DL2025-26 course resource library aims to provide structured, high-quality deep learning learning materials for global learners, addressing the aforementioned pain points.

DL2025-26 is one of the core courses in UAB's AI degree program, aiming to help students gradually master modern deep learning technologies from scratch. The course emphasizes the combination of theory and practice: students can not only learn the mathematical principles behind neural networks but also deepen their understanding through programming exercises, ultimately gaining a solid theoretical foundation and engineering ability to solve practical problems.

The course content covers four core modules:

1. Neural Network Fundamentals: Artificial neuron models, activation functions, backpropagation, etc.;
2. Convolutional Neural Networks (CNN): Principles of convolution/pooling layers, classic architectures (LeNet, AlexNet, ResNet, etc.), and residual connection technology;
3. Recurrent Neural Networks (RNN): LSTM/GRU variants, sequence modeling, and attention mechanisms;
4. Transformer: Self-attention mechanism, multi-head attention, positional encoding, and concepts of pre-trained models such as BERT/GPT.

The course is equipped with numerous programming assignments and projects, implemented using Python and PyTorch/TensorFlow frameworks. The task difficulty progresses step by step, from linear classifiers to building image recognition systems. In addition, the course focuses on cultivating debugging and optimization abilities, teaching how to diagnose problems such as gradient vanishing and overfitting, as well as their solutions, to help students handle practical projects with confidence.

The open-source GitHub repository contains resources such as course notes, programming assignments, datasets, and reference implementations, with a clear structure that facilitates self-study. Due to its open-source nature, learners worldwide can participate in discussions and form an active learning community. Both beginners and intermediate learners can gain value from it, especially providing self-learners who cannot attend formal degree programs with a free opportunity to access quality close to that of university classrooms.

After completing the course, students will have the ability to develop and deploy deep learning systems, and have strong employment competitiveness in fields such as computer vision and natural language processing. At the same time, the critical thinking and problem-solving abilities cultivated by the course help learners continue learning in the rapidly developing AI field and adapt to new technological changes.

UAB's open-sourcing of the DL2025-26 course reflects the responsibility of higher education institutions to promote knowledge sharing. Whether you are a computer science student, a career-changer engineer, or an AI enthusiast, it is worth investing time in learning. Mastering deep learning technology will be the key to seizing opportunities in the AI era. It is recommended that learners make full use of GitHub resources, actively participate in community discussions, and combine theory with practice to improve their skills.