In the study of plankton, it is common to count and identify them manually with the use of a microscope and sampling containers, which can be a tedious process. To address this problem, a Python-based neural network will be created to automatically identify common phytoplankton genera in Possession Sound. Since the most abundant phytoplankton in Possession Sound are diatoms, which include Thalassiosira, Coscinodiscus, and Chaetoceros, the network’s primary purpose will be to identify these genera. The neural network will be trained using approximately 1000 photos of each genus in varying orientations and lighting conditions, with the images being drawn from research trips aboard the Ocean Research College Academy vessel Phocoena beginning in 2007. After completing the training process, the network’s performance will be validated using samples taken at two sites around Possession sound, and it will be determined whether it meets a benchmark of 80% accuracy. It is expected that a number of challenges will be encountered with distinguishing between phytoplankton that are distorted or layered on top of one another, and these issues could be further addressed in the future. Despite these possible problems, the neural network shows promise as a low-cost alternative to current automated phytoplankton identification devices such as the FlowCAM, which can cost upwards of $100,000.

The magnitude at which a minimal change in initial conditions can affect data results was first observed by Edward Lorenz in 1963. Through his examination of chaos emerged the discovery that many natural systems are governed by chaotic behavior. The main complication of chaos theory is that nonperiodic unstable systems are unpredictable, and therefore many natural systems have yet to be understood because of its complexity. Our research considers the unstable nature of the Lorenz attractor and its influence on a center of gravity. By examining intervals of data and comparing their centers of gravity, we found that as the amount of data points tends to infinity, a center of gravity never converges to a single point. We also examine what appears to be a stark regularity and group of symmetries under an extension M(n+k)=f(Mn) of Lorenz’ original M(n+1)=f(Mn) study of Poincare sections Z.

In 2007, David Vokoun et al. derived a formula for the force of interaction between magnets. The formula is called the Gilbert's Model. According to the Gilbert’s Model, the force between two ferromagnets is given by a constant factor proportional to the saturation magnetization of each magnet multiplied by a function of the separation distance and geometry of the magnets. We show that the assumed constant is better described as a function of hyperbolic tangent of the separation distance due to the effects of magnetic field interactions on the magnetizations of each magnet, and we demonstrate that the inclusion of a simple toy 1D Ising model acting as a perturbation on the background magnetizations better predicts magnetic coupling of cylindrical magnets over small distances.