| Line 1: | Line 1: | ||
| − | '''Review 1''' | + | '''Review 1 - summary''' |
| − | * | + | * In this slecture, the author explains how to generate Gaussian random numbers from two categories with different priors. The author presented Gaussian random numbers generation for both 1D and 2D space using MatLab script. During the lecture, the author shared the code and MatLab program opened on screen and executed data synthesis and plotted its statistics on live. For the 1D example, the author utilized MatLab built-in function ("random" and "randn") to generate data and tried different combinations of parameters to see its functionality. The results were visualized using a histogram and a pie chart. The histogram showed how the density of each class looked like while the pie chart illustrates the ratio between both class 1 and class 2, which is directly related to prior probabilities. For the 2D example, the author used MatLab built-in function "mvnrnd" to generate multi-dimensional Gaussian random numbers. In the second work, the author explained a covariance matrix which is a different thing compared to the 1D example. a scatter plot and a pie chart were methods to visualize the results. From all results displayed in the lecture, it can be concluded that such methods work well for generation of Gaussian random number for multi-dimensional space. |
| − | '''Review 2''' | + | '''Review 2 - strengths''' |
| − | * | + | * This slecture explains each step of Gaussian random numbers generation in detail so that readers who do not have background knowledge about probability and random process can easily follow and generate desired outputs. Explanation with screen sharing helps readers get clear understanding of how it works, what the functions for, why such parameters are needed. Also the visualization technique the author chose is well-suited for the delivering the purpose of this slecture. I can easily catch changes and read relations by looking at the plots. |
| − | '''Review 3''' | + | '''Review 3 - suggestions''' |
| − | * | + | * Whereas this slecture serves as a step-by-step tutorial for students, some might want to see more advanced topics such as complexity of each built-in sampling function, which method they use, how long it takes to generate large samples, and how accurate the sampled distribution is. I think covering analysis on efficiency, complexity and accuracy would make this slecture more concrete. |
written by Jonghoon Jin. | written by Jonghoon Jin. | ||
Revision as of 19:38, 30 April 2014
Review 1 - summary
- In this slecture, the author explains how to generate Gaussian random numbers from two categories with different priors. The author presented Gaussian random numbers generation for both 1D and 2D space using MatLab script. During the lecture, the author shared the code and MatLab program opened on screen and executed data synthesis and plotted its statistics on live. For the 1D example, the author utilized MatLab built-in function ("random" and "randn") to generate data and tried different combinations of parameters to see its functionality. The results were visualized using a histogram and a pie chart. The histogram showed how the density of each class looked like while the pie chart illustrates the ratio between both class 1 and class 2, which is directly related to prior probabilities. For the 2D example, the author used MatLab built-in function "mvnrnd" to generate multi-dimensional Gaussian random numbers. In the second work, the author explained a covariance matrix which is a different thing compared to the 1D example. a scatter plot and a pie chart were methods to visualize the results. From all results displayed in the lecture, it can be concluded that such methods work well for generation of Gaussian random number for multi-dimensional space.
Review 2 - strengths
- This slecture explains each step of Gaussian random numbers generation in detail so that readers who do not have background knowledge about probability and random process can easily follow and generate desired outputs. Explanation with screen sharing helps readers get clear understanding of how it works, what the functions for, why such parameters are needed. Also the visualization technique the author chose is well-suited for the delivering the purpose of this slecture. I can easily catch changes and read relations by looking at the plots.
Review 3 - suggestions
- Whereas this slecture serves as a step-by-step tutorial for students, some might want to see more advanced topics such as complexity of each built-in sampling function, which method they use, how long it takes to generate large samples, and how accurate the sampled distribution is. I think covering analysis on efficiency, complexity and accuracy would make this slecture more concrete.
written by Jonghoon Jin.
