BUSS 1651
Student’s Name
Institutional Affiliation
Course Name and Code
Professor’s Name
Submission Date
, Task 1 – Effect of Activation Function (Linear)
1.1 Circle Linear
Figure 1. Circle linear
The network, which was used with the data of the Circle dataset and linear
activation, required about 274 epochs and a training loss of about 0.498 and a test loss of
about 0.503. The choice line is virtually linear and does not take into account the
structure of rings and huge quantities of points located on the outer and inner circle are
incorrectly categorized. It shows that not all patterns that are non-linearly separable, e.g.
concentric circles, can be well modeled using a pure linear model.
Student’s Name
Institutional Affiliation
Course Name and Code
Professor’s Name
Submission Date
, Task 1 – Effect of Activation Function (Linear)
1.1 Circle Linear
Figure 1. Circle linear
The network, which was used with the data of the Circle dataset and linear
activation, required about 274 epochs and a training loss of about 0.498 and a test loss of
about 0.503. The choice line is virtually linear and does not take into account the
structure of rings and huge quantities of points located on the outer and inner circle are
incorrectly categorized. It shows that not all patterns that are non-linearly separable, e.g.
concentric circles, can be well modeled using a pure linear model.
