Week8: Aug 2 - Aug 8After getting a basic idea on how the code is working, I started to run some experiments by changing some of the parameters in the code for the algorithm. The code partitions the dataset based on the specified attributes and creates a histogram. The code tests the performance of the algorithm by computing absolute count and relative count. The code is differntially private and uses Laplace noise to achieve differential privacy based on a specified parameter called the privacy budget. I ran some experiments by changing the privacy budget parameter and analyzed the results. I also found a bug in the code while running experiments. There is a section in the code that generates a specified number of random queries to obtain absolute count and relative count. The problem was that the code was not taking into account all the random queries generated to obtain the count values. I fixed the bug by making sure that the code takes into account all random generated queries while obtaining count. Thus, now if there are 1000 random queries generated the code will account all 1000 queries to obtain the absolute count and relative count values.Besides running experiments on the code, I also read a paper on Privacy Integrated Queries (PINQ). PINQ is a interactive platform for privacy-preserving data analysis and uses differential privacy to achieve privacy. I have had a great time exploring Atlanta over the past few weeks. This city is just amazing especially the downtown Atlanta area. Interestingly, many large corporations have their headquarters in Atlanta! This weekend was just more exploring and relaxing and trying new restaurants. |