This document summarizes a presentation comparing the RMSE and MAP metrics for evaluating CNN and LSTM models. It finds that RMSE is effective for measuring the accuracy of continuous predictions from CNNs and LSTMs, while MAP excels at assessing performance on tasks requiring precise retrieval. The presentation describes the methodology, including training CNNs and LSTMs on flower and CIFAR-10 datasets. Results show RMSE and MAP values after 100 epochs, with MAP generally lower. It concludes that understanding the complementary nature of RMSE and MAP enhances model assessment, and the right metric depends on the specific task.