This week I have been working on my presentation and have refined my proposal a bit. I will keep working on my presentation until the upcoming Wednesday as I wait for feedback on my second draft of my proposal
This week I have worked on the presentation. That’s it.
This week I worked on the final proposal. Feedback from the second draft indicated that there were some grammar and structure errors. I added the testing and abstract sections. I hope to finish the proposal by the end of this week.
I read some new papers and research about different modeling algorithms and started to worry about the accuracy on my system. The accuracy is not only rely on the modeling but also based on the dataset for training and the quality of acoustic input (the speaking environment). But selecting a suitable modeling algorithm is important. Now the popular models are: HMM, VQ, DTW, GMM, UBM, i-Vector. I temporarily chose hybrid GMM-UBM. I might change in the future or mix other modeling to enhance the accuracy. My goal is to reach an accuracy at least 90%.
This week, I continued working on my project proposal, submitting my second draft after some much-needed updates. I still need to work further on the Related Works section. I additionally continued working on early implementation of the project. Lastly, I prepared a first draft of my presentation slides.
I have mostly worked on the second draft of my proposal in the past week. I simplified and restructured my framework model. I also looked for more use cases of my project and decided what frameworks to use while implementing it.
This week I have not been able to do much progress. I have decided to scrap the idea to use Machine Learning in the module for altering images, due to difficulty in implementation. Besides that, I have worked on the second draft of my proposal.
This week, I worked on the similar project posted online. While working on it, I found some challenges in modifying the content-based data set to fit the collaborative-filtering method. I might end up modifying my project from a hybrid recommender to a content-based recommender. But I will keep looking for alternatives to make it possible.
During this week I met with Charlie to review feedback for the first draft of the proposal. I reconstructed this version for the 2nd draft. I also read some papers for the next pass. These were more directly related to the software I am going to use.
I found an API for food calories. I worked on proposal second draft. I explored an image editing software I am thinking to use.
Still working on the second draft. Rebuilding the diagram of SLMS. Also ready with the draft for the presentation.
I realized that the database I found for food nutrition is not enough for the calorie estimation so I looked for more resources. I started learning SQL.
I discussed my proposal draft with my advisor. I got her feedback and suggestion, and knew how to revise and improve my proposal. In the past week, I read more papers about the GMM-UBM modeling method that I plan to use for my project. I understood the specific procedure now but it is still hard to fully understand this principle… Now my another problem is to find a suitable dataset and decide if my system is text-dependent. There are three primary ways for speaker verification now: text-dependent, mixed, text-independent. The text-independent way is very difficult and complicated to do because user can say anything to pass the verification. But text-dependent way is restricted and not safe for spoofing attacks. For example, people can replay pre-recorded voice to pass the verification. Therefore, the mixed way is better. It restricts the text in a way but safe for spoofing attacks. For example, they user can only speak numbers one – ten, but every time the text is random. But it is hard to find a dataset of all audio file in numbers in English. Now I need to decide which text way my system will use.
This previous week, the work I’ve done has been two-pronged, as has become the norm and will continue to be for the rest of this semester. First, I continued work on the basic implementation of the game. I currently have the control module working, as well as a looping stage that I created in order to test the controls. On the proposal side, I’ve been making edits based on the in-class peer review that we did, as well as working more recently based on the feedback given by Xunfei. I also met with Xunfei to go over her feedback of my first draft, and updated her on my progress.
This week, I investigated the technologies being used in my found papers more closely to find which technologies would be more feasible for my project. For data collection, I have found that the facebook-sdk python library (https://pypi.org/project/facebook-sdk/) used by Pool and Nissim is the best option to connect to the Facebook Graph API, since it looks well documented and has all the options I might need. I also decided to use the Facebook Pages of politicians as my dataset. I reread As the Tweet, So the Reply?: Gender Bias in Digital Communication with Politicians by Mertens et al. to see if their methods could be adapted to my project. I will need to look at their references for methods in more detail to see if I can feasibly apply them to my project.
This past week, I worked on mainly reading my new papers. I did a third pass reading on all my old papers and did at least second pass reading on the new ones. I tried finishing more than half of the existing project on python notebook and played with the data set. I now have a better sense of how to start my project next semester. I also met Xunfei and updated my progress to her. As soon as the feedback for proposal draft 1 comes out, I will be revising my writing and finishing the existing project I have been working on. I also plan to test the existing project on collaborative filtering to make sure it works with a different data set.
In the past week, I have spent most of my time working on the first draft of the proposal. I decided to research and include a new category of papers in my proposal that I had not spent a lot of time before on. The new category that I included was “Sentiment Analysis.” While working on the proposal and refining the design of my framework, I realized that sentiment analysis, something that has been thoroughly covered by researchers of neural networks is very close to my research since I also need to know the sentiment behind the email/piece of text that is to be improved.
This week I have
This week I have:
In the past week, I have used my peer review from Jordan, as well as my own proof-reading of a physical copy of my draft to fix a lot of errors. I wrote my draft in a bit of a rush, and as a result, there were a lot of formatting errors, most of which I have now fixed. I have also updated some of my diagrams in accordance with feedback I have received and expanded some content in my draft that needed to be clarified.
In addition to working on my draft, I have been working on my project itself (preliminary work can be found in the git repo). I created a mockup GUI to give me some ideas about how I want to design the actual version next semester, as well as testing some implementations of different filters, operators, and edge detectors. Some of these results will hopefully be represented in the next version of my draft.
During the past week, I finished the first draft of my proposal and started to make those changes for the second draft. I have also continued reading some papers for their next pass. I continued to watch videos and read content related to the USB Rubber Ducky. I have started to put together some scripts that I would like to use for the attack. I also spoke with Charlie to refine my methods for the physical attacks I am going to implement. I now have a better/ more related CS implementation for this attack than what I previously had. During this next week, I am going to be working more with Metasploit on Kali Linux.
For this week I looked into many different datasets, including gis fire map data https://frap.fire.ca.gov/mapping/gis-data/ and Kaggle dataset https://www.kaggle.com/elikplim/forest-fires-data-set but couldn’t find what I was looking for.
I picked The Ranch Fire in California but couldn’t find good datasets for it. I was trying to find I’m trying to find elevation, wind direction, humidity, and vegetation.
All of them have to contain coordinates so I can layer them together. Also, I need each set at different time stamps for the simulation. I will discuss this during the next weekly meeting.
For this week I have done the following tasks:
Learned to view shapefile contents with Netlogo and Python library pyshp. A shapefile is an Esri vector data storage format for storing the location, shape, and attributes of geographic features. This type of file is quite complicated so it took me sometimes to understand the format and its contents.
I had trouble finding the right dataset for my project. Charlie suggested that I look into https://www.frames.gov/afsc/partners/fmac/guides-products. I downloaded the data for Alaska but it does not have the contents that I was looking for. Finding the right data is currently a big challenge.
I am also getting more familiar with Netlogo. Using Netlogo, I could view the content of the data for Alaska and also extract the metadata using the command line provided with NetLogo.
I have been continuing learning machine learning with Python, specifically PyTorch. I started with PyTorch because it has a less steep learning curve compared to Tensorflow (the alternative). However, there are more tutorials for Tensorflow and I might pivot next semester as the image processing gets more complicated and I need more resources in incorporating image processing into the machine learning. I think I will be able to build both a ResNet and AlexNet algorithm and compare them to decide which one to use. I have also written the code for video editing in Python to convert the input video into frames. For this task I am using OpenCV. It is straightforward to do this. I have not yet decided how many frames I will take in the first round.
Discussed the feedback on the first draft with Xunfei. Got valuable feedback and planning to implement them in the second draft. I am searching for the algorithms to be used in the SLMS and working on the second draft.
Finished propsal first draft. Explored softwares for image processing.
Finished my first draft of proposal. I read some blogs about speaker verification tech and found out that I was wrong on some aspects (actually I was confused). Those blogs help me understand more and deeper about speaker verification. So I revised my framework and flowcharts: take voice input -> feature extraction -> modeling -> database. The modeling part is the most difficult part in speaker verification. The most popular models are: Hidden Markov Model, Gaussian Mixture Model, Vector Quantization, etc. I am not sure which one I will use for sure. It all depends on my dataset and customer need. I need to experiment several models to know which one I want the best. But I chose GMM temporarily on my proposal.
I only worked on finishing the first draft of the proposal this week.
This past week, I’ve finalized the basic design for the game I will be implementing. It will be a horizontal auto-runner, where the player ducks/jumps to avoid obstacles to the beat of the music in order to keep playing. I continued familiarizing myself with Unity2D, and plan on starting work on the game this upcoming week. Additionally, I wrote up the first draft of my project proposal.
This past week, my work has been split in two directions: First, I’ve been refamiliarizing myself with Unity, by means of going through my Game Design second project. Further than that, I’ve been familiarizing myself with Unity2D for the first time, which I plan on using for the senior project due to the simplicity as compared to Unity3D. Besides getting used to the main software engine I will be using, I also continued reflection on my proposal outline; I’ve been looking more into different PCG-G algorithms and have decided on using the chunk paradigm as my second stage generation algorithm. Its stages won’t be as directly aligned to the music, but it should improve efficiency.
I extensively worked on the proposal last week, reading more papers and writing out what I plan to do helped me figure out the scope of the proposed project. I also experimented a bit more with tensorflow. I made some changes to my initial framework design to now include a frontend and backend for the end-user to interact with.
This week, I revised my diagram again and selected papers to use for my proposal. I read through all of them carefully and wrote an outline for my proposal. I discussed the details of my ideas with Xunfei, modified some, and finished writing the proposal.
Much of my work this week has consisted on working through the first draft of the proposal as well as reading some papers for the next pass. I also found a couple other sources to use. These new sources were not research papers but rather articles related to my project.
I finally received the Gourmet Dataset. In fact, I received a devised version that has twice as many images as the original one. I also have the Yelp dataset, although that dataset has not be curated by humans, I am hoping to use it for training my algorithm in addition/instead of ImageNet or AVA.
Since I already have gotten access to the datasets, I have been reading about ResNet/AlexNet implementations, which was my goal for next week.
This week I met with Charlie to discuss my project design. We also talked about GIS extension, which is a library to handle GIS data for NetLogo. Charlie talked about how to layer different types of data on a base map. The most important tasks for the upcoming weeks are to figure out how to find different types of data for a fire location and how to process the data.
I worked extensively on finishing the first draft. I rewrote the design and related works section.