deep learning notebook

.gitignore

@@ -0,0 +1,22 @@
+__pycache__/
+.ipynb_checkpoints/
+*.pyc
+.env
+.venv/
+venv/
+.DS_Store
+.vscode/
+
+data/*.csv
+data/*.parquet
+data/*.db
+data/03_hold_difficulty/*.csv
+data/04_climb_features/*.csv
+data/05_predictive_modelling/*.csv
+data/06_deep_learning/*.csv
+data/06_deep_learning/*.npy
+models/*.pth
+models/*.pkl
+models/*.csv
+
+BLOG.md

README.md

@@ -62,7 +62,7 @@ The utility [`BoardLib`](https://github.com/lemeryfertitta/BoardLib) is used for
 We'll work with the Tension Board 2. I downloaded TB2 data as `tb2.db`, and I also downloaded the images.
 
 ```bash
-# install boardlib
+# install boardlib (also in requirements.txt)
 pip install boardlib
 
 # download the database

data/05_predictive_modelling/model_summary.txt

@@ -0,0 +1,35 @@
+
+### Model Performance Summary
+
+| Model | MAE | RMSE | R² | Within ±1 | Within ±2 | Exact V | Within ±1 V |
+|-------|-----|------|----|-----------|-----------|---------|-------------|
+| Linear Regression | 1.467 | 1.882 | 0.782 | 42.6% | 73.3% | 34.9% | 79.4% |
+| Ridge Regression | 1.467 | 1.882 | 0.782 | 42.6% | 73.3% | 34.9% | 79.4% |
+| Lasso Regression | 1.475 | 1.891 | 0.780 | 42.2% | 73.0% | 34.6% | 79.3% |
+| Random Forest (Tuned) | 1.325 | 1.718 | 0.818 | 47.0% | 77.7% | 38.6% | 83.0% |
+
+### Key Findings
+
+1. **Tree-based models remain strongest on this structured feature set.**
+   - Random Forest (Tuned) achieves the best overall balance of MAE, RMSE, and grouped V-grade performance.
+   - Linear models remain useful baselines but leave clear nonlinear signal unexplained.
+
+2. **Fine-grained difficulty prediction is meaningfully harder than grouped grade prediction.**
+   - On the held-out test set, the best model is within ±1 fine-grained difficulty score 47.0% of the time.
+   - The same model is within ±1 grouped V-grade 83.0% of the time.
+
+3. **This gap is expected and informative.**
+   - Small numeric errors often stay inside the same or adjacent V-grade buckets.
+   - The model captures broad difficulty bands more reliably than exact score distinctions.
+
+4. **The project’s main predictive takeaway is practical rather than perfect.**
+   - The models are not exact grade replicators.
+   - They are reasonably strong at placing climbs into the correct neighborhood of difficulty.
+
+### Portfolio Interpretation
+
+From a modelling perspective, this project shows:
+- feature engineering grounded in domain structure,
+- comparison of linear and nonlinear models,
+- honest evaluation on a held-out test set,
+- and the ability to translate raw regression performance into climbing-relevant grouped V-grade metrics.

data/06_deep_learning/neural_network_summary.txt

@@ -0,0 +1,33 @@
+
+### Neural Network Model Summary
+
+**Architecture:**
+- Input: 119 features
+- Hidden layers: [256, 128, 64]
+- Dropout rate: 0.2
+- Total parameters: 72,833
+
+**Training:**
+- Optimizer: Adam (lr=0.001)
+- Early stopping: 25 epochs patience
+- Best epoch: 121
+
+**Test Set Performance:**
+- MAE: 1.270
+- RMSE: 1.643
+- R²: 0.834
+- Accuracy within ±1 grade: 49.0%
+- Accuracy within ±2 grades: 80.2%
+- Exact grouped V-grade accuracy: 39.2%
+- Accuracy within ±1 V-grade: 84.3%
+- Accuracy within ±2 V-grades: 96.8%
+
+**Key Findings:**
+1. The neural network is competitive, but not clearly stronger than the best tree-based baseline.
+2. Fine-grained score prediction remains harder than grouped grade prediction.
+3. The grouped V-grade metrics show that the model captures broader difficulty bands more reliably than exact score labels.
+4. This makes the neural network useful as a comparison model, and potentially valuable in an ensemble.
+
+**Portfolio Interpretation:**
+This deep learning notebook extends the classical modelling pipeline by testing whether a neural architecture can improve prediction quality on engineered climbing features.
+The main result is not that deep learning wins outright, but that it provides a meaningful benchmark and helps clarify where model complexity does and does not add value.

models/feature_names.txt

@@ -0,0 +1,119 @@
+angle
+total_holds
+hand_holds
+foot_holds
+start_holds
+finish_holds
+middle_holds
+is_nomatch
+mean_x
+mean_y
+std_x
+std_y
+range_x
+range_y
+min_y
+max_y
+start_height
+start_height_min
+start_height_max
+finish_height
+finish_height_min
+finish_height_max
+height_gained
+height_gained_start_finish
+bbox_area
+bbox_aspect_ratio
+bbox_normalized_area
+hold_density
+holds_per_vertical_foot
+left_holds
+right_holds
+left_ratio
+symmetry_score
+hand_left_ratio
+hand_symmetry
+upper_holds
+lower_holds
+upper_ratio
+max_hand_reach
+min_hand_reach
+mean_hand_reach
+std_hand_reach
+hand_spread_x
+hand_spread_y
+max_foot_spread
+mean_foot_spread
+foot_spread_x
+foot_spread_y
+max_hand_to_foot
+min_hand_to_foot
+mean_hand_to_foot
+std_hand_to_foot
+mean_hold_difficulty
+max_hold_difficulty
+min_hold_difficulty
+std_hold_difficulty
+median_hold_difficulty
+difficulty_range
+mean_hand_difficulty
+max_hand_difficulty
+std_hand_difficulty
+mean_foot_difficulty
+max_foot_difficulty
+std_foot_difficulty
+start_difficulty
+finish_difficulty
+hand_foot_ratio
+movement_density
+hold_com_x
+hold_com_y
+weighted_difficulty
+convex_hull_area
+convex_hull_perimeter
+hull_area_to_bbox_ratio
+min_nn_distance
+mean_nn_distance
+max_nn_distance
+std_nn_distance
+mean_neighbors_12in
+max_neighbors_12in
+clustering_ratio
+path_length_vertical
+path_efficiency
+difficulty_gradient
+lower_region_difficulty
+middle_region_difficulty
+upper_region_difficulty
+difficulty_progression
+max_difficulty_jump
+mean_difficulty_jump
+difficulty_weighted_reach
+max_weighted_reach
+mean_x_normalized
+mean_y_normalized
+std_x_normalized
+std_y_normalized
+start_height_normalized
+finish_height_normalized
+start_offset_from_typical
+finish_offset_from_typical
+mean_y_relative_to_start
+max_y_relative_to_start
+spread_x_normalized
+spread_y_normalized
+bbox_coverage_x
+bbox_coverage_y
+y_q25
+y_q50
+y_q75
+y_iqr
+holds_bottom_quartile
+holds_top_quartile
+angle_x_holds
+angle_x_difficulty
+angle_squared
+difficulty_x_height
+difficulty_x_density
+complexity_score
+hull_area_x_difficulty