# Modeling Paper Roadmap

## Modeling techniques survey

### CPU modeling:

1. Energy of Computing using Performance Monitoring Counts (PMC):
- Three ways to measure: System-Level with external measurements, on-chip power sensors and energy prediction models.
- Dynamic and static analysis.
- PMC with positive. correlation.
- Selection of relevant PMC by Pearson correlation matrix.

2. Regression models for performance and power prediction:
- Polynomial regressors.
- Predictors/Variable merging.
- Splines or Piecewise. polinomials to fit measurements.
- 12 set of predictors with multiple variables.

### GPU modeling:

1. HALWPE Performance estimation for GPUs:
- Ensembles of 12 linear models and one non linear.
- Optimizatin on LSE (like in system identification).
- Non-linearity comes from random forest.
- Validation using 10 fold validation like in ML.
- Error minimization to choose best model.
- Not tested on energy consumption.

### FPGA modeling:

1. CMOS FPGA modeling
-