AI_ASS_2 notes

Notes:

• Working on dataset given CC.csv data
  • row = applicant
  • col=info

For x=row#, y=appApproved?, $w_j$ = $\pm 1$ unknown @param attempting to find, then:

$$f(x)=w_1{x}^1+w_2{x}^2+\cdots +w_6{x}^6$$

Where $f(x)$ represents some linear relationship btwn the inputs →result. Suppose w is some int array (hence must only contain -1/1), then we want to find the combo of -1’s and 1’s that converge to ??

We can define $er(w)$, where $|w|=6$ as:

$$er(w):=\frac{1}{n}\sum _{i=1}^{340}(f(x_i)-y_i{)}^2$$

Where 340 is our row length, then each round for some random combo of w we have:

Dataset changes

• CreditApprove, 1=approve
• Gender: M=1
• CarOwner: Y=1
• PropertyOwner,Y=1
• Children=#
• WorkPhone, Has?=1
• Email has? =1

Problem 1)

• Hill Climb, minimize er(w) =w
• Examine all solutions adj to cur solutionthen id best

will need to plot:
x-axis = round of search
y-axis = er(w)
Must see that er(w) Converges