The Siemens Competition in Math, Science & Technology was founded by the Siemens Foundation in 1999, and has since become one of the most popular research competitions for high school students in the nation. Every year, several students from Ward Melville’s InSTAR program apply to Siemens after conducting research the summer before their senior year. On October 16, 2014, seniors Niyati Desai, Gary Ge, and Sapna Nath were informed that they were semifinalists.

When asked what her reaction was when she received the news, Niyati Desai said, “I was really surprised. I had no idea. I did a lot of work at the last minute for this. I had finished my paper early, but I had to edit it multiple times based on revision suggestions from people, so I had to put in a lot of effort at the last second. But I think it ended up really helping.” She then commented on her decision to submit to the competition, saying, “When I was submitting to Siemens, I was kind of on the fence about it, since it isn’t mandatory. People were telling me I should probably focus on other things, since there are so many things going on fall of senior year, but I had this amazing experience over the summer, and I wanted to learn from it, so I really encourage all juniors who have an interest in science to do research, and you should submit to competitions because you really feel proud of your work, you feel like you did a good job, and you worked hard for it.” Below is a brief description of her project:

Silicon technology is reaching limitations leaving little room for further improvement in speed, size and power efficiency. I studied whether graphene-ferroelectric hybrid devices could be a viable alternative to silicon based technology which is widely used in everything from radios to computers. Graphene, a 2-dimensional layer of carbon with uniquely useful electrical properties, was deposited on top of carefully aligned ferroelectric layers, which intrinsically have 2 possible charge orientations that correspond to “1”s and “0”s, the common form of memory storage. In my research, the requirements for memory storage were fulfilled at low temperature and my results showed that these devices can potentially serve as silicon replacements.

Since he knew that the chances of getting to be a semifinalist were very slim, Gary Ge was also surprised when he was received the news. In an effort to get the results early, Gary took advantage of an option that Siemens provided that allowed students to find out the results via social media a couple of hours before the official release. Gary expressed his anticipation and excitement by “frantically stealing all of [his] friends’ phones to get onto the Facebook app” to find out whether or not he made it to the next round. He offered advice to student researchers regarding the Siemens competition: “Siemens, unlike Intel, has a very, very early deadline, so I highly recommend getting started early with your paper. Also, don’t be afraid to not do it if it means compromising your health. It’s tough, but if you tough it out, you might be rewarded with the result.” Below is a brief description of his project:

Images and other visual media contain a tremendous amount of information, and recent studies in computer vision have sought ways develop algorithms that automatically derive this information for image understanding. Of particular interest is the problem of recognizing actions that humans in an image appear to be performing, a task that has numerous applications ranging from automated annotation of large datasets like ImageNet or Flickr to realtime action detection in security footage. However, despite tremendous progress made in the past few years, action recognition remains a problem that can only be reliably performed by humans. In order to close the gap between computer and human performance, I analyze human gaze patterns and combine them with a state-of-the-art action classification algorithm.

I utilize a subset of images provided by the PASCAL VOC actions dataset and evaluate eye movement data that was collected from subjects tasked with recognizing actions in these images. This data is visualized in a number of ways that allow for the elucidation of spatial, temporal and durational patterns in gaze. Patterns are then quantified to create several novel gaze features, while features for a state-of-the-art method are obtained as a baseline. Using these two sets of features I train two Support Vector Machines, which are supervised learning methods that classify elements according to previously learned patterns, and use them to classify images into ten different action classes. The confidence outputs from both classifiers are then combined with weights in order to illustrate how gaze can improve upon compute vision algorithms.

This study finds that a classifier trained with features using gaze patterns can perform comparably well with one trained with computer vision features at a much lower computational cost. When their outputs are combined I show that gaze data has the potential to improve on the performance of state-of-the-art methods.

Sapna Nath was very happy when she found out the results. She joked, “My fingers were really happy too – since there were some days when they really started to hurt from typing.” She offered advice to students interested in the research field: “Enjoy what you are doing and do research to learn something new. Research is not a single competition – it is just the start of your own unique journey in science, so it’s okay to make mistakes. The most important thing is that you learn something new and understand what you are doing. Also, try not to procrastinate, and take really detailed notes during the summer.” Below is a brief description of her project:

The Fas Receptor is a protein contained in cells of our body. When this receptor is stimulated correctly it causes the cell to die or grow depending on the receptor’s location and distribution within the cell. By modifying the structure of the Fas Receptor in kidney cancer cells, I was able to determine the exact part of the receptor that is responsible for the distribution. Knowing what region controls the distribution of the receptor and consequently whether the cell will die or grow, enables us to investigate new ways to prevent cancer using the receptor’s cell signaling mechanisms.