diff --git a/Running Start/AD450 - Data Science Development/Discussion - Looking into the Fundamentals of Machine Learning.md b/Running Start/AD450 - Data Science Development/Discussion - Looking into the Fundamentals of Machine Learning.md
index 1e517c5..481c459 100644
--- a/Running Start/AD450 - Data Science Development/Discussion - Looking into the Fundamentals of Machine Learning.md	
+++ b/Running Start/AD450 - Data Science Development/Discussion - Looking into the Fundamentals of Machine Learning.md	
@@ -4,4 +4,6 @@
 
 The article that I chose to look at is an introduction to the log-loss score and some intuition behind it: https://towardsdatascience.com/intuition-behind-log-loss-score-4e0c9979680a/
 
-The goal of this article is to show some of the intuition between log-loss rather than just providing formulas and showing how to apply them. It starts by defining classification problems and how models first predict a probability that an instance belongs to a certain class before then classifying them based on the probability. It then describes how log
\ No newline at end of file
+The goal of this article is to show some of the intuition between log-loss rather than just providing formulas and showing how to apply them. It starts by defining classification problems and how models first predict a probability that an instance belongs to a certain class before then classifying them based on the probability. It then describes how log-loss is a metric that is 0 when the model perfectly predicts the class and goes higher when it diverges more from the actual value. It also discusses how log-loss grows exponentially rather than linearly to punish predictions that are further away from reality more. It then shows the formula for log-loss and how it is calculated for a model. Lastly, it talks about some of the limitations including how log-loss can be misleading on imbalanced datasets.
+
+I have seen log-loss used as a metric to evaluate more basic machine learning models but it can also be used for more complicated models like neural networks. This can be important since it allows us to tune the models and get them to be more accurate than they might have been otherwise. The example used in the article used is classifying spam emails and something like log-loss can be useful to punish models that predict incorrectly more and train them to more accurately classify spam emails and reduce scams. It is useful to know about different metrics so that I know what I can use in the future and some of the benefits and downsides to each to make the best decision.
\ No newline at end of file