A curated collection of selected projects developed for the Pattern Recognition course at Warsaw University of Technology (WUT) during the 2025–2026 Erasmus+ academic year.
The repository covers three complementary approaches to classification:
- probabilistic classification
- linear multiclass classification
- artificial neural networks
The strongest projects in the repository are the linear-classification and neural-network tasks, which focus on MNIST recognition and progressively more advanced learning methods.
Folder:
bayesian-classification/
Classical statistical pattern-recognition methods, including:
- Bayes classification
- independent Gaussian models
- multivariate Gaussian models
- Parzen-window estimation
- 1-NN classification
- feature-selection experiments
Folder:
linear-classification/
Multiclass handwritten-digit recognition using perceptron-based ensembles.
Topics include:
- One-vs-One classification
- One-vs-Rest classification
- PCA
- feature expansion
- voting and rejection
- confusion matrices
- MNIST classification
Folder:
neural-network-classification/
Artificial neural-network classification implemented in MATLAB / GNU Octave.
The project compares:
- baseline backpropagation
- normalized-input training
- momentum-based training
- confusion matrices and error reports
pattern-recognition/
├── bayesian-classification/
├── linear-classification/
├── neural-network-classification/
├── .gitignore
└── README.md
The MNIST binary dataset is intentionally not stored in the repository.
The linear and neural-network projects expect:
train-images.idx3-ubyte
train-labels.idx1-ubyte
t10k-images.idx3-ubyte
t10k-labels.idx1-ubyte
inside their respective directories.
This keeps the repository lightweight while preserving reproducibility.
- MATLAB
- GNU Octave
- Pattern Recognition
- Machine Learning
- Bayesian Classification
- Perceptrons
- Artificial Neural Networks
- PCA
- Support for MNIST data
- Confusion-Matrix Analysis
Developed for the Pattern Recognition course at Warsaw University of Technology (WUT) during the 2025–2026 Erasmus+ academic year.
Selected tasks:
- Task 2 — Bayes Classification
- Task 3 — Linear Classification
- Task 4 — ANN Classification
Álvaro Buendía Senise
- GitHub: alvaarobuendia
- LinkedIn: Álvaro Buendía Senise