docs: add line about future validation

docs/final-report/methods.Rmd

@@ -18,4 +18,4 @@ Once the refinement process is done, the refined separation vector is accepted b
 
 A further improvement to the algorithm that we we did not have time to implement would be a re-learning feature. The user would run the algorithm on a sample of the data, and based on that sample identify firing times that are inaccurate. Then they would run the algorithm on the rest of the data and based on the firing times identified as inaccurate, the algorithm  would no longer make similar mistakes in the rest of the decomposition. Implementing this feature would be quite complex because it is unclear how this would be done. One idea was that we somehow change the initialization of the separation vectors so that they no longer identity the false firing times when applied to the pre-processed data. However, since the separation vector changes throughout the LCA and refinement processes, it would be hard to control the effect that it actually has on the estimated firing times. Another approach would be to influence the KMeans algorithm so that the threshold for the small peaks cluster includes the peaks that were  as false firings, in the hopes that future peaks of similar size are also false firings. The downside to this approach would be that we may increase the amount of incorrect identifications of large peaks as small peaks, which are discarded.
 
-The stakeholders affected by our blind source separation algorithm are researchers and those that would be affected by their research. This is why it is important our algorithm works properly so that researchers results are accurate and do not affect the general public adversely down the line. For example, if someone uses `EMGdecomPy` and obtains inaccurate results, and these results are used to inform a neuromuscular diagnosis down the line, it could greatly affect someone's life. We have not had the chance to thoroughly validate our algorithm, as the debugging process took a great deal of time. We have only received qualitative results, obtained by  visually comparing MUAP shapes identified by `EMGdecomPy` and those @Hug2021 identified using the `DEMUSE` tool. There are concerns with this approach as the `DEMUSE` software uses a similar but different algorithm than @negro_muceli_castronovo_holobar_farina_2016. The `DEMUSE` software is a highly validated software in comparison to `OT Bioelettronica`, and therefore the partner wishes to compare our results to theirs.
+The stakeholders affected by our blind source separation algorithm are researchers and those that would be affected by their research. This is why it is important our algorithm works properly so that researchers results are accurate and do not affect the general public adversely down the line. For example, if someone uses `EMGdecomPy` and obtains inaccurate results, and these results are used to inform a neuromuscular diagnosis down the line, it could greatly affect someone's life. Periodically, the results of our algorithm should be compared to others to obtain a second opinion on the decomposition of the EMG signal. We have not had the chance to thoroughly validate our algorithm, as the debugging process took a great deal of time. We have only received qualitative results, obtained by  visually comparing MUAP shapes identified by `EMGdecomPy` and those @Hug2021 identified using the `DEMUSE` tool. There are concerns with this approach as the `DEMUSE` software uses a similar but different algorithm than @negro_muceli_castronovo_holobar_farina_2016. The `DEMUSE` software is a highly validated software in comparison to `OT Bioelettronica`, and therefore the partner wishes to compare our results to theirs.