Ever since I was a little boy, the world of mathematics has always intrigued me. It soon became my favorite subject when I started school as I found myself enjoying the process of solving complicated problems and exploring the reasoning behind mathematical concepts. While pursuing mathematics as my undergraduate major, however, I found that I wanted to challenge myself in an area where I could utilize my skill of math on a more practical platform. Over the past two years, I have felt the most earnest sense of fulfillment when utilizing my mathematical methodologies and computational techniques to solve some interesting deep learning problems like crime prediction and object detection. I have realized machine learning and artificial intelligence …show more content…
Supervised by Professor Andrea Bertozzi and Dr. Bao Wang, our task was to create a functional object detector to distinguish police and people other than police from the body worn video cameras provided by the LAPD. Instead of just applying our dataset and parameters to an existing network as I did in the previous research, this time we were required to understand the inherent network structure of Faster Rcnn since we needed to add some extra convolutional layers to fit our specified detection target. Based on our mathematical knowledge about stochastic gradient descent and optimization, we implemented a new loss function by combining center loss and contrastive loss, and applying the gradient boosting method to improve detection accuracy. After 2 months of hard work, we successfully increased the detection accuracy by 1.6 percent compared to benchmark data. From this summer research program, I have mastered some advanced topics like the forward and backward propagation in neural networks and modification of convolutional layers to achieve our own purpose. Besides being more enthusiastic towards computational mathematics and machine learning, I also found some interesting topics I want to further explore. For instance, because our detection object shared …show more content…
Besides providing a strong mathematics and programming foundation in your breadth requirements, your program also has some specific and interesting electives for machine learning track like Deep Learning for Computer Vision and Natural Language Processing. The faculty here and the research opportunities of your program also attract me so much. I am particularly interested in Professor Chang’s research in visual relation detection. The novel idea of his research inspires me a lot and it will be a great pleasure for me to have an opportunity to join the DVMM lab. In addition, Columbia University also provides a strong connection with industry. Besides your central location in New York, Columbia’s computer science department has attracted so many successful technology companies which provides me the best chance to explore cutting-edge ideas and grasp career opportunities. I believe that the systematic curriculum and the vigorous research environment of your program can not only deepen my knowledge foundation but also give me a broadened insight in the advancing fields of artificial intelligence and machine