OloriBern/learning-tutor-embeddings-weight-predictor-elastic

---
library_name: sklearn
tags:
  - sklearn
  - regression
  - embeddings
  - weight-prediction
  - elastic
model-index:
  - name: Elastic Net
    results:
      - task:
          type: regression
          name: Embedding Weight Prediction
        metrics:
          - type: mse
            value: 0.7620290676891099
            name: Test MSE
          - type: r2
            value: -0.044365932705230184
            name: Test R²
---
# Elastic Net Weight Predictor

Linear model combining L1 and L2 regularization

## Performance Metrics

- Training Time: 1.55 seconds
- Training MSE: 0.739313
- Testing MSE: 0.762029
- Training R²: 0.003906
- Testing R²: -0.044366

## Model Analysis

### Predictions vs True Values
![Predictions](./plots/predictions.png)

This plot shows how well the model's predictions match the true values:
- Points on the red line indicate perfect predictions
- Spread around the line shows prediction uncertainty
- Systematic deviations indicate bias

### Error Distribution
![Error Distribution](./plots/error_distribution.png)

This plot shows the distribution of prediction errors:
- Centered around zero indicates unbiased predictions
- Width shows prediction precision
- Shape reveals error patterns

### Dimension-wise Performance
![Dimension MSE](./plots/dimension_mse.png)

This plot shows the MSE for each embedding dimension:
- Lower bars indicate better predictions
- Variations show which dimensions are harder to predict
- Can guide targeted improvements

## Usage

```python
import skops.io as sio

# Load the model
model = sio.load('weight_predictor_elastic.skops')

# Make predictions
weights = model.predict(question_embedding)
```