Jump to content

Recommended Posts

Posted

I am a freshman Computer Science student who is applying for a prestigious research internship at IST Austria. This is my SOP:

 

I am applying as an intern at the Lampert Group for Computer Vision and Machine Learning at IST Austria.

Being aware that the best way to learn about any field is to start working on it, I decided to take up a project on Machine Learning. My journey as a Machine learning newbie began with this project. After some thorough searching, I thought about solving the problem of classification on unbalanced datasets. It is difficult to implement classification algorithms on these datasets as there is a danger of overlapping of classes. Solving this problem led me to this paper on Mahalanobis Data-Oversampling. It uses Mahalanobis distance as a metric to measure the covariance in place of the the euclidean distance. It also made me realise the importance of machine learning in solving real life problems like detection of anomalies, fraud detection in banking transactions, etc.

 

Success in this project led me to another project on an interesting problem. I took up a project on the problem of audio classification. Substantial work had been done on a similar problem of image classification. Here, deep learning and convolutional neural networks succeeded where traditional techniques had previously failed. Some of these networks even passed human classifiers in accuracy. Audio classification brought with it a different set of interesting challenges. SVMs and deep convolutional networks continued to dominate the scene here too.

 

Initially, I solved a simpler single label classification problem on the Google Audioset. Taking cue from the results obtained, I tried solving a similar problem for multiple-labels. During this, I learned about many multi-label classification algorithms and also about the various problems in multi-label classification. This also gave me the required exposure to a variety of problems like incorrect labelling, computationally expensive applications, etc. This also cleared my brain of misconceptions about having a lucky breakthrough in science. The goal of research is not to publish but to make an impact to the field. It is important to fail in research and learn from the failures. It also helped me on my journey from a machine learning newbie to a machine learning enthusiast.

 

During my work on the audio classification problem, my professor recommended me to use Zernike descriptors for the project. Studying about Zernike desciptors helped me cultivate a different approach towards my problem and introduced me to Image Processing. Zernike moments help in giving an invariant transform of images and play a big role in Computer Vision too. Learning more about Zernike descriptors led me to study their applications in 3D object retrieval and other fields like bioonformatics too. I also became aware of the interesting problems in Computer vision like learning from continuous data. One of the most important work to solve this problem is being performed at IST Austria.

 

In my whole journey, I realised the importance of having an excellent peer group and teachers. Lampert Group has been involved in solving some of the most intriguing and difficult problems of computer vision. Working there will help me learn from some of the most esteemed professors in this field and also help me progress in my journey from a machine learning enthusiast to a machine learning student.

 

 

This is the first time ever that I have written a statement of purpose. I would love to receive your feedback regarding the same.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...

Important Information

This website uses cookies to ensure you get the best experience on our website. See our Privacy Policy and Terms of Use