Beginning Anomaly Detection Using

gradient, which can be 

Figure 3-11.  The formula for the mean squared error cost function

Figure 3-12.  A formula for the mean squared error cost function, with more 

specific notation separating the weights and the biases

Figure 3-13.  An elaboration on what the function h(w, b) means in Figure 

3- 12

Chapter 3   IntroduCtIon to deep LearnIng

---

83

thought of as similar to the derivative but in a multi-dimensional setting, and adjust the 

weights in a direction that would change the gradient so it approaches 0.

There are several optimization algorithms to help the model achieve the optimal 

weights including