Alejandro Treny Ortega · Applied statistics, machine learning and scientific computing
A portfolio of 22 studies across 18 project directories, combining coursework and exploratory experiments in Python, R and C++. Each project includes its methods, code and interpretation; the project guides below identify the data and how to reproduce the analysis.
Execution guide · Review and results · Dataset provenance
| Study | Focus | Tools |
|---|---|---|
| Dry Bean Classification | Six classifiers, training cross-validation and prediction explanations | Python, scikit-learn, Quarto |
| Student Performance | Exploratory associations and a decision tree with a majority baseline | R, caret, R Markdown |
| RNA-Seq Classification | PCA pipelines and four classifiers on a five-class tumor dataset | R, caret, Quarto |
| Kepler PCA | Standardized principal components and interpretation of astronomical measurements | R, Plotly, Quarto |
| Kepler MDS and Clustering | Mixed-type distances, multidimensional scaling and cluster diagnostics | R |
| RGB Projection Pursuit | Fisher separation, whitening and sequential/parallel orthogonal projections | R, OpenImageR, foreach |
| Music Evolution | Playlist metadata and lyric-derived summaries from cached snapshots | Python, pandas, TextBlob |
| Release-Decade Prediction | Random-forest classification of balanced Spotify audio features | Python, scikit-learn |
| Study | Focus | Tools |
|---|---|---|
| Linear Programming and Absolute-Error Regression | Resource balances, duality and regression as a linear program | Python, Gurobi, Plotly |
| Mixed Integer Production Planning | Fixed costs, production tiers and logical constraints | Python, Gurobi, Quarto |
| Matrix Chain Ordering | Dynamic programming, RPN execution and OpenMP benchmarks | R, RcppArmadillo, C++ |
| Deep Q-Trading | Chronological Bitcoin evaluation with explicit trading costs | Python, PyTorch, Gymnasium |
| Generative Adversarial Networks | DCGAN training and qualitative evaluation on MNIST | Python, PyTorch |
| Neuroevolution | Genetic optimization of LunarLander policies with separate test seeds | Python, Gymnasium |
| Study | Focus | Tools |
|---|---|---|
| Epidemic Dynamics | Stochastic cellular automata and spatial scenario comparisons | Python, NumPy, SciPy |
| Physarum Simulation | Particle sensing, trail deposition, diffusion and animation | Python, NumPy |
| Monopoly Monte Carlo | Landing frequencies and simplified cash-flow scenarios | R, tidyverse |
| Gray–Scott Patterns | Reaction–diffusion dynamics on a periodic grid | Python, NumPy, Matplotlib |
| Visual Steganography | RGB payload embedding, bit-depth trade-offs and lossless recovery | Python, NumPy, Pillow |
These three experiments are collected in the CERN-inspired portfolio.
| Study | Focus | Tools |
|---|---|---|
| Anomaly Detection and Quantization | Autoencoders on synthetic jets and quantized parameter export | Python, TensorFlow |
| Neural Importance Sampling | Normalizing-flow proposals checked against a known integral | Python, PyTorch |
| Hybrid Quantum Edge Classification | Classical and simulated quantum classifiers on one TrackML event | Python, PyTorch, PennyLane |
Start with RUNNING.md for the tested Python 3.12, R 4.5.3 and Quarto environments, installation commands and project execution steps. The guide and audit are in Spanish; project READMEs are in English. Download a committed HTML report and open it locally to use its interactive plots.
Small datasets are included; larger inputs are downloaded or supplied separately as described in each project. Gurobi requires a suitable licence. Spotify/Genius credentials are needed only for optional music-data enrichment.
The 19 September 2026 review records the reproduced results, 15 verification groups and 20 dataset snapshots, together with the specific limits of each check. Results describe the documented data, splits and simulation assumptions; individual project guides explain what would require further validation.