Displayed here are the two theses I wrote at Dartmouth College.
Computer Science Thesis: “A logical re-conception of neural networks: Hamiltonian bitwise part-whole architecture.”
Abstract
Contemporary neural networks such as Convolutional Neural Networks (CNN)’s rely on a constrained set of linear-algebra premises. We describe a novel Hamiltonian Logic Network (Hnet) which derives from simple principles of binary logic operations, with an aim of working toward an architecture of neuro-symbolic AI. The Hnet performs classification and recognition and constructs part-whole relations and hierarchies using the intrinsically low-cost architecture of binary nodes and bit-wise arithmetic operations. By compositing these relationships using Hamiltonian hierarchies, we are able to create a sparse, non-adjacent tiling of any space. The Hnet’s capabilities are demonstrated on two classical datasets: the MNIST characters dataset and UCI’s German Credit dataset.
Middle Eastern Studies Thesis: “Hacking Orientalism: Theorizing Arabic Digital Writing”
Abstract
When two Arabic letters are unable to meet, the impossible happens. Join them on a journey to find themselves and love in the digital age. Watch as they encounter poorly-written English code that pretends to know everything about them without being able to differentiate between them! Look on in horror as every major design suite and word processor has an irredeemable flaw in its rendering of Arabic Text! Enjoy a thrilling techno-ride as the writer parallelizes the Missionary printing presses of the Nahda with modern word processors through a comparative reading of Ahmad Faris al-Shidyaq’s saucy postmodern text Leg over Leg! Stay until the satisfying conclusion which dissects bug reports of Arabic script! Coming to a lecture hall near you.