week of Feb 27th

Feb 23rd
yesterday I scheduled my road test for march first at 8:30AM. I really hope that I will pass this test and get my license. If i do then "the studnet has surpassed the teacher" because Cade does not have his full or even provisional license yet. Today in class I got anaconda working and installed tensorflow in a virtual enviorment with the correnct version of python. To do this I used the command "pip install tensorflow". Though the computer when through the procsess of installing the library when I ran the file I recieved the error "tesorflow not found." I am pretty confused and do not know why it is not working.

Feb 25th
today I took a differet approch to getting tensorflow to work. This time in uninstall python on the comupter then reinstalled the later version(3.7) so that it was tensorflow compatible. Though I recived the same error that tensorflow could not be found. When I tried typing in "import tensorflow as tf" directly into the terminal I recived and error that I think means that there was something wrong with the GPU. I tried going though some online solutions to solve them but none worked.

weekly summary
This week was kinda depressing because I want to finally get the model to work but the computer simply isnt finding the tensorflow libary. Next class I plan on importing the files that were returned by the earlier part of the code to my google drive then running the neural network in google colab. after I get the trained neural network file I will save it to the computer then run the rest of the code using the notepad++ and terminal method.


week of Feb 20th weblogs

Wensday Feb. 17th
Today I decided to migrate my efforts on the machine learn app to new computer. This time, instead of using google colab, I am using the terminal and notepad++. Andy helped me set it up and we worked together to troubleshoot errors with missing libaries and other things. Though I got the part of the code the repackages the data into numpy arrays working, I ran into issues on the part of the code where the the images are modified to be the right shape and split into training and test data.

Friday Feb. 19th
Today I ran into issues getting the computer to start. Thus the rest of the class was spent trying to help Will and Nick fix the fans on the computer.
By the end of class everything was fixed. I stayed alittle bit after class to try working on the program alittle more. I got the part of the code the failed before to work this time. Next the part of the code is to implement will be the actual neural network. Though as I tried to import the tesorflow library I found out that the version of python that I have is actaually "too new" for tensorflow to support. So then I figured out that I had to install anaconda so I can create a virtual enviorment where I can run an old version of python. I had trouble using powershell to get the hash value for the anaconda installation and this is about where I stopped.

Weekly summary
This week I made the switch from google colab (a machine learning platform that operates almost exculsively out of the cloud) to the text editor and terminal method. There are several pros and cons to making the switch. On one hand google colab allows me to work from any computer but on the other hand it is very inconsistant and slow. With the alternative method I get much more consistant results and can get more help from Andy and Will since they are more familiar with this method.


Feb. 13th

Tuesday, February 9th
Today I worked on trying to get the MatPlotLib rendering of the dataset to work. I got help from Will but despite our combine efforts it did not work out.

Thursday February 11th
Today i spent the class cleaning my laptop using the Ifix it kit and some canned air. I found some lose magnets inside the computer so I spent some time putting them back into place. Now when I close or shake my computer there is no noise.

Weekly summary
This week i had alot of issues with google colab. On the first run though of the code the matplotlib would produce images of the data being used, but on subsequent run thoughs of that same code block it would fail without changing anything. I suspect it has something to do with how google colab compiles code. Also it took an unessarly long time to run the code. I am thinking of switching to a different method.


Weblogs Feb. 6th

Tuesday Feb. 2nd,
today I didnt get much done because my computer died and the charger that I found wasnt powerful enough. though I did set up a workspace on a computer in the montiering lab that I can work on in case something like this happens again.

Friday Feb 5th,
practice driving and read manual in preperation for my test on the 18th. Still working on improving the accuracy of my machine learning model.

weekly summary
This week I didnt get much done with regards to my machine learning project for various reason. Next week my goal is to get a accuracy that is similar to that of the guy in the TowardsDataScience article that I am referancing.


Weblogs Jan 30th

Monday, Jan. 25
Today I help out Anuhea and Alec with the solar panel analysis. we used the mango server to look at the effeciency and total power output of the panels.

Wensday, Jan 27
Today I pitched my project to Dr. Bill. As I talked about in the last weblog my project is making an app that uses machine learning to identify the japanese characters. I worked on how to upload the data to the google colab notebook that I am using. Before I was trying to upload the files invidually to the server using the files.upload() command but this proves too time consuming and I would have to redo it each time I open the notebook. Instead I connected the colab notebook to my google drive using the drive.mount() command. From there the code could properly acess the files.

Friday, Jan 29th
Today I worked on training the katakana model. However I have alot of trouble getting the model's accurcy to imporve. I also when the model reached about the 8th epoch each time I got and error because the computer wanted to "Restoring model weights from the end of the best epoch." I am still working and resolving these errors.

weekly summary
after talking with Dr. Bill I have looking into the possibility to using this technology to teach people new to japanese how to write Japanese. I think this is totatlly possible. Another thing I was thinking of doing was having the model learn as a people new to japanese learned. Though this is something I am not sure how to do.


Weblogs Jan. 23

Tuesday Jan. 19th-
today most of the class was spent talking with Dr. Bill. At the end of class I look into some of the prospective projects and worked one a machine learning model in tensorflow.

Thursday Jan 21st-
I decided today that I would help out with the solar panel stuff. Though i spent most of my time working on how to use the function model.evalute() to test my machine learning model. It seems that the data that I was giving the model didnt have the right shape but i didnt know how to change the shape.

weekly summary-
After thinking about the best project for this semester this weekend I figured out what I want to do. I want to make an app that helps me with my japanese home work. How this app would work is that you would draw the japanese character on the phone screen using your finger then the phone would use machine learning to find the various definitions of the character. After doing some reasearch I should a article on TowardDataScience.com where there is a indepth guide on how to build said app. Though I will be using mostly using his code in my project, my contribution will be the following:

- trying different layer stuctures
- using different activation functions (the part of a model that makes neural networks non-linear)
- trying different optimizers (How the model assesses failure)
- using different regulaiztion techniques (How to keep the model from overfitting)
- expirimenting with different data augmention techniques (kind of like adding water to lemon concentrate to make lemonade)
- adding fetures to the app
- impelmenting my app in my japanese class


weblog interview


Dec. 5 weblogs

Nov 30th,
today worked on my machine learning project. since I have not made as musch progress on my former project as I would have hope Im am now just focusing on making a simple classification model.

Dec 2nd,

Today I spent all class talking story.

Dec 4th

today I practiced on the driving simulator. specifically I practiced driving curvy roads.

Weekly summary:
this week I have redirected the direction of my project. I want to have something to show during our final presentation so I am just redireting my efforts towards a different easier project.


Nov 28 weekly

Tuesday, Nov 24
Today I worked on finishing up the machine learning course. I am now looking into where to start with my next project. Because not all the the information for instagram statistics is avalible, I am worried that my model may nor workout very well. As a result, I am planning to make the best model I can in a short amount of time then move on to a more feasable project.

weekly summary
this week we only had one class so I really didnt get that much done. But this weekend I drove around a parking lot with my mom practicing parking and turning. It was really healpful having all the practice on the dirivng simulatior.


Nov 20 weekly

Tuesday, Nov 17
Today I sepent most of my time on the driving simulator. I learned how to park and turn around properly. I also expiremented with different kinds of cars to get the feel for them.

Friday, Nov 20
Today I spent most of my time workign on finishing the machine learning crash course. I learned about different metrics to measure a model. Accurcy = # correct predicitions/ # total predictions. Recall = # positive predicitons/ # correct predictions. Precision = # correct positive predictions/ #positive predictions.

weekly summary
this week I learned alot about machine learning and made it through the bulk of the course. I think that what I learned about accuracy, precision, and recall will be very useful as I continue in my project. Accurcy, while it many sound like a good metric, is often misleading. For instance, if you had a model that predicited wheather a person had a very rare diease and if it guessed all negative then its accuracy woudl be pretty good. But it would be an essentially useless model. Precision and Recall are better metrics but are in conflict with each other. when the precision raises the recall falls.


Nov 14 weekly

Monday, Nov 9th
Today I practiced driving on the simulator.

Wensday, Nov 11th

Today I split my time between doing my machine learning course and practicing driving. Most of the class I working on the machine learning stuff. Now I am using a new driving simulator that will better represent how I will actually be driving.

Friday, Nov 13th

I spent most of my time today practing on the driving simulator. Since we are using this new program, a lot of the time was spent working out technical issues with the simulator.

Saturday, Nov 14th
Weekly summary
This week I spent most of my time on the driving simulator. Though I did some machine learning during class, the amount of time that I spent in class was not enough. Going forward I will probable start devoting more time to my machine learning project.


Nov 7 weekly

This week I spent most of my time at ISR on the driving simulator. I think this experience will be very useful when I get my license in January. In my free time I have also been working on the machine learning crash course. Going forward I would like to strike a balance between spending time on the driving simulator and spending time working on my machine learning project.


Nov 5 daily

Now that the driving simulator is set up and installed on the new computer, I practiced driving. I am slowly learning how and when to shift gears.


Nov 3 daily

Today I helped Cade set up the driving simulation. I thought it would be a really cool idea to learn to drive manual on a simulator where I can make many mistakes without severe repercussions.


Oct 31 weekly

This week I've been really busy with finishing up my questbridge application so I haven't had much time to work on my machine learning project. Though I have been working on the crash course a little in my free time, there is still substantial material for me to cover. Next week will have more time to devote to completing the machine learning crash course.


Oct 30 daily

I worked on my questbridge supplements again. I only have 2 days left so I really just want to polish what I have done before I turn them in.


Oct 28 daily

I worked on my college supplements today. Since I am applying to colleges through Questbridge (link: https://www.questbridge.org/) I have supplements for 10 schools due on Nov. 1st. If I get accepted through this program I get to go to college for free.


Oct 26 daily

Today I worked on the machine learning crash course. Link: https://developers.google.com/machine-learning/crash-course. So far it seems to be a very clear and interactive way of learning machine learning.


Oct 24 weekly

All this week I worked on learning how to make an API. I learned alot about terminal commands for windows, virtual environments, what PATH variables are, and the general functions of an API. Now that I understand what APIs are and how they work I will look into APIs already in existence and tweak them to my purposes. Next I will begin working on the google machine learning crash course on tensorflow so I can begin to create a model.


Oct 22 daily

Finished up the youtube tutorial on python APIs today. Link: https://www.youtube.com/watch?v=GMppyAPbLYk. Going forward it will begin the google machine learning crash course.