Project Structure

This project is structured as follows:

• scripts: Python scripts used for conducting analysis and generating results.
• Data: Contains the model used in our experiments.
• results: Output from scripts, for sample set analysis.
• test_results_mar: Output from scripts, for test set analysis.
• MAP_MAR_data.csv: Contains metrics such as MAP/MAR for various datasets with differing configurations of exponent bits and mantissa bits. Used in Figures 4, 7, and Table 1.
• AAI_mult: chisel code avaliable for building hardware
• floating_mult: chisel code avaliable for building hardware
• floating_add: chisel code avaliable for building hardware

Recreating Data

1. Generate MAP/MAR data with different exponent bits, mantisssa bits for Figures 4, 7, and Table 1:

   sbatch run.sh
   

2. Resampling:

   sbatch run_sample.sh 
   

3. replacement for Figure 6 and Table 4:

   sbatch run_replacement.sh 
   

4. figure8: use 'numerical_loss_before' for plot of before error correction

   python3 analysis.py
   

5. figure9: use data in folder test_results_mar

   python3 analysis.py