Author: Keyin Wu

Posted on

NOC – Week 6 Assignment – Keyin Wu (Quoey)

Here is the link to my assignment 6.

It is consisted of two moving rectangles and a couple of balls. The two rectangles move up and down and every time it collides with the border, it will change the color. For the balls, their colors are based on the collision with other balls as well as the rectangles. If they collide with other balls, the rgb value will turn into half of the sum of the other ball and itself, as if there is a passing effect. If they collide with the two rectangles, they will immediately change into the rectangle’s current color. When the number of the ball is enough, there will be some gradient effect on screen where the colors are also change all the time. Besides, the balls have attraction between each other. In order to control its speed to avoid them moving too fast, I also manipulated the parameters when they collide with each other.

Furthermore, if the user presses the mouse, their velocity will be reduced and not recover util the user releases the mouse. And I did this by changing the framerate because I didn’t find a better way to manipulate their velocities directly. 

 

Posted on

IML | Week06 Midterm Concept Proposal- Quoey Wu

Midterm Concept — The Weight of Stroke
Introduction

For my midterm concept, I’m inspired by the balance. When we talk about the balance, it is often connected with the comparison. But things on the two sides of the balance tend to be in the same dimension, and it is hard for human’s brain to reach a conclusion between objects in different dimensions, for example, a sentence and a piece of drawing. However, I think it might be not difficult for machine by calculation. As a result, I would like to build a balance model where two users can input some visual created by canvas by round. Then the machine will calculate the weights of these strokes and show the balance’s status corresponding to the result. I think this model reflects the conversion between abstract comparison to some concrete visual effect which to some degree explores the boundary of human’s mind.

Posted on

IML | Week05 Cifar-10 Model- Quoey Wu

At first, I started with 100 epochs. But the process went too slow so I quit it and restarted with 3 epochs.

Then I tried different batch sizes starting from 2048 to 32. I found the runtime didn’t change too much and was around 6-7 minutes in each case. As for the test loss and accuracy, here are the results.

  • batch_size = 2048

 

  • batch_size = 1024

 

  • batch_size = 512 

 

  • batch_size = 256

 

  • batch_size = 32

So, we can see as the batch size gets smaller, the test accuracy gets higher with the test loss decreasing. 

Then I kept batch size fixed at 32 and tried different epochs number. I found that when the epochs number gets bigger,  the test accuracy also gets bigger.

As a result, I think the parameters in the training process are fundamental to the result. In order to get more accuracy, we may need to decrease the batch size and increase the epochs. Also, since the growth of epochs will increase the runtime, it also needs some limit to reduce the runtime in a reasonable range. But I am not very clear about the principle behind it, which I would like to explore more in the future.