Applied Deep Learning: A Case-Based Approach to Understanding Deep Neural Networks

© Umberto Michelucci 2018

Umberto MichelucciApplied Deep Learninghttps://doi.org/10.1007/978-1-4842-3790-8_6

6. Metric Analysis

Umberto Michelucci¹ 

(1)

toelt.ai, Dübendorf, Switzerland

 

Let’s consider the problem we analyzed in Chapter 3 for which we performed classification on the Zalando dataset. While doing all our work, we made a strong assumption without explicitly saying it: we assumed that all the observations were correctly labeled. We cannot say that with certainty. To perform the labelling, some manual intervention was needed, and, therefore, a certain number of images were surely wrongly classified, as humans are not perfect. This is an important revelation. Consider the following scenario: in Chapter 3, we achieved roughly 90% accuracy ...