Open Access. Powered by Scholars. Published by Universities.®
Physical Sciences and Mathematics Commons™
Open Access. Powered by Scholars. Published by Universities.®
Articles 1 - 7 of 7
Full-Text Articles in Physical Sciences and Mathematics
Methods For Efficient Object Categorization, Detection, Scene Recognition, And Image Search, Alessandro Bergamo
Methods For Efficient Object Categorization, Detection, Scene Recognition, And Image Search, Alessandro Bergamo
Dartmouth College Ph.D Dissertations
In the past few years there has been a tremendous growth in the usage of digital images. Users can now access millions of photos, a fact that poses the need of having methods that can efficiently and effectively search the visual information of interest. In this thesis, we propose methods to learn image representations to compactly represent a large collection of images, enabling accurate image recognition with linear classification models which offer the advantage of being efficient to both train and test. The entries of our descriptors are the output of a set of basis classifiers evaluated on the image, …
Cutting Wi-Fi Scan Tax For Smart Devices, Tianxing Li
Cutting Wi-Fi Scan Tax For Smart Devices, Tianxing Li
Dartmouth College Master’s Theses
Today most popular mobile apps and location-based services require near always-on Wi-Fi connectivity (e.g., Skype, Viber, Wi-Fi Finder). The Wi-Fi power drain resulting from frequent Wi-Fi active scans is undermining the battery performance of smart devices and causing users to remove apps or disable important services. We collectively call this the scan tax problem. The main reason for this problem is that the main processor has to be active during Wi-Fi active scans and hence consumes a significant and disproportionate amount of energy during scan periods. We propose a simple and effective architectural change, where the main processor periodically computes …
Shared Roots: Regularizing Deep Neural Networks Through Multitask Learning, Piotr Teterwak
Shared Roots: Regularizing Deep Neural Networks Through Multitask Learning, Piotr Teterwak
Dartmouth College Undergraduate Theses
In this paper, we propose to regularize deep neural nets with a new type of multitask learning where the auxiliary task is formed by agglomerating classes into super-classes. As such, it is possible to jointly train the network on the class-based classification problem AND super-class based classification problem. We study this in settings where the training set is small and show that , concurrently with a regularization scheme of randomly reinitializing weights in deeper layers, this leads to competitive results on the ImageNet and Caltech-256 datasets and state-of-the-art results on CIFAR-100.
Constant Rmr Transformation To Augment Reader-Writer Locks With Atomic Upgrade/Downgrade Support, Jake S. Leichtling
Constant Rmr Transformation To Augment Reader-Writer Locks With Atomic Upgrade/Downgrade Support, Jake S. Leichtling
Dartmouth College Undergraduate Theses
The reader-writer problem [1] seeks to provide a lock that protects some critical section of code for two classes of processes: readers and writers. Multiple readers can have access to the critical section simultaneously, but only one writer can have access to the critical section to the exclusion of all other processes. The difficulties in solving the reader-writer problem lie not only in developing a correct and efficient algorithm, but also in rigorously formulating the desirable properties for such an algorithm to have. Bhatt and Jayanti accomplished both of these tasks for several priority variants of the standard reader-writer problem …
Chain Match: An Algorithm For Finding A Perfect Matching Of A Regular Bipartite Multigraph, Stefanie L. Ostrowski
Chain Match: An Algorithm For Finding A Perfect Matching Of A Regular Bipartite Multigraph, Stefanie L. Ostrowski
Dartmouth College Undergraduate Theses
We consider the problem of performing an edge coloring of a d-regular bipartite multigraph G = (V, E). While an edge coloring can be found by repeatedly performing Euler partitions on G, doing so requires that the degree of G be a power of 2. One way to allow the Euler partitioning method to continue in cases where d is not a power of 2 is to remove a perfect matching from the graph after any partition that results in a graph with an odd degree. If this perfect matching can be identified in O(E) time, we can maintain the …
Stylecheck: An Automated Stylistic Analysis Tool, Alexander P. Welton
Stylecheck: An Automated Stylistic Analysis Tool, Alexander P. Welton
Dartmouth College Undergraduate Theses
StyleCheck is a user-friendly tool with multiple functions designed to aid in the production of quality writing. Its features include stylistic analysis (on both document-wide and individual-sentence scales) and spelling and grammar check, as well as generating suggested replacements for all types of errors. In addition, StyleCheck includes the capability to identify the famous author (out of a limited corpus) with the style most similar to the user's. The source code for StyleCheck is available online at: https://github.com/alexpwelton/StyleCheck
Modflows: Methods For Studying And Managing Mesh Editing Workflows, Jonathan D. Denning
Modflows: Methods For Studying And Managing Mesh Editing Workflows, Jonathan D. Denning
Dartmouth College Ph.D Dissertations
At the heart of computer games and computer generated films lies 3D content creation. A student wanting to learn how to create and edit 3D meshes can quickly find thousands of videos explaining the workflow process. These videos are a popular medium due to a simple setup that minimally interrupts the artist’s workflow, but video recordings can be quite challenging to watch. Typical mesh editing sessions involve several hours of work and thousands of operations, which means the video recording can be too long to stay interesting if played back at real-time speed or lose too much information when sped …