Hey everyone,
Been a long time since I posted.
Last summer I had a paid internship / paid summer research position. I was handed two gigantic textbooks the first day of the job. One said "Machine Learning with Python" and the other "Data Mining." These textbooks were massive and had unbelievably complicated math in them. I thought, "Wow I am so far in over my head. I am beyond underqualified and have no clue what I am doing, or how I got this job."
Fast forward...I was to develop a script for the Computer Science department at Hamilton College. This script (python code, machine learning / data mining) would take previous Neuroscience experiments conducted at Syracuse University (and future experiments) and apply "Machine Learning" on the data. The experiments conducted used fNIRS (read about it here: https://en.wikipedia.org/wiki/Functional_near-infrared_spectroscopy) to figure out oxygenated/deoxygenated blood levels in parts of the brain. The goal was to record brain states of human beings (when you are feeling suspicious, trustworthy, nothing at all, etc).
At any rate, I created, with only one other person, a Python script (.py) to convert millions upon millions of numbers into fewer numbers that made more sense to researchers and the data mining program called "Orange." Finally, machine learning was used so that predictions could be made based upon the data. Essentially, all subject data was recorded in the experiment, a computer was "trained" on the data, and then it attempted to predict the "condition" (brain state) of a novel subject. This was one of many, many, many, many ways the script would predict brain states. There were many more ways.
Each method of machine learning / predicting was done across many, many subjects, and hundreds of result excel documents (that I designed) were produced--all by this one Python script. There were countless methods of predictions used, and thankfully algorithms exist (check them out here https://en.wikipedia.org/wiki/List_of_machine_learning_concepts) or else this would never be possible. Between the "control" and "experimental" groups, the accuracy was 100%. As in, a computer could tell with 100% accuracy whether a subject's experimental brain data was either in the control condition or not. However, it was less accurate at trying to predict the correct condition between the experimental groups, but still fairly accurate and hoping to improve with more subject data.
The script itself, once written by myself and 1 other Computer Science student, took 3 weeks total to run, 24 hours a day (so in total about 504 hours estimated), and used 6-12 different processors simultaneously. I did not use multi-threading. In addition, Python is a slow language. Originally, they wanted to use a super computer with 100s of processors, but it just did not pan out that way. Eventually though, all of the results were generated! Much of this ran while I had already left from the premises, and so I checked everything remotely every so often and posted the final results to a mutual dropbox.
I tried to sum this all up -- it's quite complex. I definitely did not write in this post even a fraction of what I did. At any rate, I'm hoping the experiments + results will be published by summer of 2016. I was told that I might have to fly out to California for a conference, but I think just the head researchers will be going. If everything goes as planned, I will be one of four to eight co-authors for each of two published papers.
What impressed everyone was that I was the youngest person, with the least amount of coding experience, out of anyone. I had only one "Intro to Computer Science" (language: Python) class under my belt. The people I worked for could not comprehend that I was pre-med and not going into the field of computer science and technology. They said they had never met an individual like me ever before, and could not believe the amount of results I generated with only one other person.
Well, I'm here to say that I finally am going to change my course in life. I'm currently studying at Vanderbilt University, soon to finish the "Medicine, Health, and Society" (MHS) major. Next semester I will finally add Computer Science classes, double major in Computer Science and MHS, and drop the pre-med track altogether. I'm about to withdraw from Organic Chemistry, and damn that feels good.
I am finally going to be able to do what I am most passionate about. I'm finally going into the field of computers and technology. All of my life, I have experimented with computers, mobile phones, etc, doing things none of my classmates (other than the ones who are now going for Computer Science) were doing.
I'm beyond excited, and just wanted to let everyone know.
Been a long time since I posted.
Last summer I had a paid internship / paid summer research position. I was handed two gigantic textbooks the first day of the job. One said "Machine Learning with Python" and the other "Data Mining." These textbooks were massive and had unbelievably complicated math in them. I thought, "Wow I am so far in over my head. I am beyond underqualified and have no clue what I am doing, or how I got this job."
Fast forward...I was to develop a script for the Computer Science department at Hamilton College. This script (python code, machine learning / data mining) would take previous Neuroscience experiments conducted at Syracuse University (and future experiments) and apply "Machine Learning" on the data. The experiments conducted used fNIRS (read about it here: https://en.wikipedia.org/wiki/Functional_near-infrared_spectroscopy) to figure out oxygenated/deoxygenated blood levels in parts of the brain. The goal was to record brain states of human beings (when you are feeling suspicious, trustworthy, nothing at all, etc).
At any rate, I created, with only one other person, a Python script (.py) to convert millions upon millions of numbers into fewer numbers that made more sense to researchers and the data mining program called "Orange." Finally, machine learning was used so that predictions could be made based upon the data. Essentially, all subject data was recorded in the experiment, a computer was "trained" on the data, and then it attempted to predict the "condition" (brain state) of a novel subject. This was one of many, many, many, many ways the script would predict brain states. There were many more ways.
Each method of machine learning / predicting was done across many, many subjects, and hundreds of result excel documents (that I designed) were produced--all by this one Python script. There were countless methods of predictions used, and thankfully algorithms exist (check them out here https://en.wikipedia.org/wiki/List_of_machine_learning_concepts) or else this would never be possible. Between the "control" and "experimental" groups, the accuracy was 100%. As in, a computer could tell with 100% accuracy whether a subject's experimental brain data was either in the control condition or not. However, it was less accurate at trying to predict the correct condition between the experimental groups, but still fairly accurate and hoping to improve with more subject data.
The script itself, once written by myself and 1 other Computer Science student, took 3 weeks total to run, 24 hours a day (so in total about 504 hours estimated), and used 6-12 different processors simultaneously. I did not use multi-threading. In addition, Python is a slow language. Originally, they wanted to use a super computer with 100s of processors, but it just did not pan out that way. Eventually though, all of the results were generated! Much of this ran while I had already left from the premises, and so I checked everything remotely every so often and posted the final results to a mutual dropbox.
I tried to sum this all up -- it's quite complex. I definitely did not write in this post even a fraction of what I did. At any rate, I'm hoping the experiments + results will be published by summer of 2016. I was told that I might have to fly out to California for a conference, but I think just the head researchers will be going. If everything goes as planned, I will be one of four to eight co-authors for each of two published papers.
What impressed everyone was that I was the youngest person, with the least amount of coding experience, out of anyone. I had only one "Intro to Computer Science" (language: Python) class under my belt. The people I worked for could not comprehend that I was pre-med and not going into the field of computer science and technology. They said they had never met an individual like me ever before, and could not believe the amount of results I generated with only one other person.
Well, I'm here to say that I finally am going to change my course in life. I'm currently studying at Vanderbilt University, soon to finish the "Medicine, Health, and Society" (MHS) major. Next semester I will finally add Computer Science classes, double major in Computer Science and MHS, and drop the pre-med track altogether. I'm about to withdraw from Organic Chemistry, and damn that feels good.
I am finally going to be able to do what I am most passionate about. I'm finally going into the field of computers and technology. All of my life, I have experimented with computers, mobile phones, etc, doing things none of my classmates (other than the ones who are now going for Computer Science) were doing.
I'm beyond excited, and just wanted to let everyone know.
Last edited:
