Using geopandas and the Overpass API, climbing center locations were spatially joined with population density data across Spain's autonomous communities. Cataluña had the highest count (37 centers), while Madrid showed the highest population density (830 people/km²).

Part 2 of optimal control post. After establishing why open-loop control fails in part 1, I implement full Linear Quadratic Gaussian (LQG) control for coordinated eye-head movements. This combines optimal feedback control via Riccati recursion with Kalman filtering for state estimation under partial observability.

This is part 1 of a two-part series on optimal control from coursework I did in 2024. I investigate why saccadic eye movements systematically undershoot their targets - spoiler: it's actually optimal when noise scales with effort. Starting from analytical optimization of saccadic gain, I extend to constrained eye motion control with Lagrange multipliers, building toward the full LQG implementation in part 2.

Using spectrograms as input data, a CNN was trained to classify Hebrew yes/no utterances with over 90% accuracy. UMAP dimensionality reduction revealed two distinct clusters corresponding to the two words.

Poisson tiles are subdivisions of a plane where points appear according to a Poisson distribution with probability p = λ/n². The project animates three key properties: stationarity (uniform distribution under translation), independence (separate regions don't affect each other), and absence of accumulation (finite nodes in bounded regions).

Manual counting of climbing centers across Spanish metropolitan areas using Google Maps revealed significant variation in climbing center density, ranging from 27,279 to 275,154 people per climbing center. The study found Cadiz as the most dense and Torrelavega as the least dense city with centers.