During this summer, three of my family friends and I embarked on an exciting journey by enrolling in UT Dallas' AI Summer Camp. After eight weeks of immersive learning, I'm now ready to provide insights into this unique experience and help you determine if the $1200 registration fee is a worthy investment.
First up, let's talk logistics. Way before the camp began, we faced a little challenge – getting to the campus. The camp runs from 9:30 AM to 4:00 PM, and even though you can do it online, trust me, being there in person is way better. You get to learn from the teachers and understand the complex material much better. This means that a lot of people, me and my friends included, had to find ways to cut down on travel costs. We found that carpooling was an excellent solution that reduced fuel costs and allowed flexibility for our parents' schedules. So, if you're thinking of going, sort out your ride first.
Next, lets get to the main part of the camp - the material covered. When me and my friends came into this camp, we knew very little, if any, Python and were mostly experienced in Java. This entire course is taught in Python, and while we were able to pick things up by the 2nd or 3rd week, I 100% recommend trying to learn as much Python as you can before the beginning of the camp. The basic timeline for the camp was that we would start with some simple Python games and learning the math used for the course, transition into learning ML and neural networks and finally develop our own project at the end.
The first 2 weeks were focused on 3 main things: learning Python, learning the math needed and making some simple projects like a tic-tac-toe game where you can play against the computer. Near the end of this phase, things got a lot more tough, as we started to go into topics like Multivariable Calculus, Linear Regression and SVMs. This, combined with the introduction of neural networks and ML made weeks 3-5 very difficult since there were tons of big concepts being thrown at us very quickly. However, around the end of week 5, I started to get the hang of things much easier when I decided to do more research on the upcoming topics before they were taught. For this period, I definitely recommend taking good notes on the different types of neural networks such as CNNs, RNNs and LSTMs. Also, brace yourself for tough parts – it's good to be ready.
When we started to get into neural network part of the course, we got to do some cool projects. These included: a decision tree finding the best column in a data set, linearly separating data, analysing emotion in an IMDB review dataset, determining what number is in a picture and making a bot to complete an 8 puzzle. If we completed any of the projects early, we could work on extra projects the instructors would give us or play around with different machine learning topics we found interesting, with the instructors to help us. Another suggestion I have if you're worried about the upcoming final project, is to pick a algorithm that you think you'll use in your project and try to learn it fully. This will give you much more confidence for the project and for trying out the rest of the algorithms, since all of them are based off each other.
Looking back at the course, a lot of the content we learnt were college level content. While our class had a variety of experience levels, from no knowledge of Python like me to college students with multiple years of experience, I think all of us would have done far worse than we did if it weren't for our fantastic instructors. They were fantastic at condensing big topics into shorter lessons only taking up a day or two, and while they had some limitations in explaining these topics, our feedback helped them grow in teaching. They were also very engaging and fun to be around, which lessened the stress of coding for 8 hours straight. A specific thing that they did that I found was very effective was not spoon-feed us answers, but nudge us in the right direction. This method helped us a lot with our individual debugging skills, and is something I think more teachers should try using.
A great comment our instructors made was that if you can't teach a topic to a 5 year old, you haven't learnt it properly. They would often make some of us go up to the front of the classroom and teach it to our peers. While the social anxiety of this was worrying, I'll admit that we learnt best and worked our hardest when we were under this pressure. Personally, I was asked to teach LSTMs to some of my neighbours in the class, and this pressure forced me to learn the topic to heart.
This brings me to an important part of this course that people joining don't tend to think about - the mental aspect. It can be quite difficult to code for 8 hours straight for 8 weeks straight, and I noticed that a lot of people started to get burned out at various points throughout the day. I don't really have a specific recommendation for this apart from just trying to have break points throughout the day and being aware of this.
Finally, we get to our final projects. For it, we were allowed to make groups of up to 3 and present our idea to the instructors. Next, while we worked on it, we had to keep up to date a sheet covering our project updates. We didn't work on the project alone every day, as the instructors mixed it up a bit by also including lessons such as recursive learning and how to read an academic paper. At the end of the 8 weeks, we presented our project to the class.
My final project was an Exoplanet Habitability Classifier, inspired by my passion for aerospace. We used a decision tree to analyse a NASA dataset of exoplanets and find which columns were best to classify by. We then used a neural network to classify a user's input of data. In parallel, we made a GAN with CNNs to generate images of exoplanets based on another NASA exoplanet dataset, with artistic representations of these exoplanets. Finally, we used a graphical user interface to display all of these together in a visually appealing way.
Some other projects showcased were algorithms for a drone to follow a path, neural networks to detect cavities in teeth and a project that took sheet music and turned it into an audio file of the music. While some people didn't finish their project in the given time frame, I was very impressed by all the projects that were shown off, and this reflected very well on how much we had learnt due to this course.
In conclusion, I highly recommend attending the UT Dallas 8-Week AI Workshop as I believe it gives a great in-depth foundation on AI coding, can be incredibly useful in a portfolio in a world where AI is the next big thing, and is just generally quite fun. If you're interested in joining, I advise remembering my tips I've mentioned previously. I hope everyone who attends this camp in the future has a ton of fun and makes some great projects. A big thanks to the stellar teachers and organisers, and to my friends for making this so fun!
Related Links:
Comments
Post a Comment