Feature Selection from Lyme Disease Patient Survey Using Machine Learning

Abstract

Lyme disease is a rapidly growing illness that remains poorly understood within the medical community. Critical questions about when and why patients respond to treatment or stay ill, what kinds of treatments are effective, and even how to properly diagnose the disease remain largely unanswered. We investigate these questions by applying machine learning techniques to a large scale Lyme disease patient registry, MyLymeData, developed by the nonprofit LymeDisease.org. We apply various machine learning methods in order to measure the effect of individual features in predicting participants’ answers to the Global Rating of Change (GROC) survey questions that assess the self-reported degree to which their condition improved, worsened, or remained unchanged following antibiotic treatment. We use basic linear regression, support vector machines, neural networks, entropy-based decision tree models, and k-nearest neighbors approaches. We first analyze the general performance of the model and then identify the most important features for predicting participant answers to GROC. After we identify the “key” features, we separate them from the dataset and demonstrate the effectiveness of these features at identifying GROC. In doing so, we highlight possible directions for future study both mathematically and clinically.

Controllable Inversion of Black-Box Face-Recognition Models via Diffusion

“…We mainly compare our model with the three methods that generate faces from black-box features whose code is available online: NbNet [36] ("vgg-percept-nbnetb" parameters), the method by Razzhigaev et al [48] (referred to as "Gaussian sampling"), and the method by Vendrow and Vendrow [61] (referred to as "StyleGAN search").…”

“…While capturing the identity of the input face well in some cases, NbNet [36] and Gaussian sampling [48] both fail to produce realistic faces. In contrast, StyleGAN search [61] always generates high-quality images, but they are not always faithful to the original identity, sometimes failing completely (by getting stuck in local minima) as seen in the last row. Our method is the only method that produces high-quality, realistic images that consistently resemble the original identity.…”

“…Original image NbNet [36] Gaussian sampling [48] StyleGAN search [61] ID3PM (Ours) Figure 3: Qualitative evaluation with state-of-the-art methods. The generated images of our method (with InsightFace [8] ID vectors) look realistic and resemble the identity of the original image more closely than other methods.…”

“…The generated images of our method (with InsightFace [8] ID vectors) look realistic and resemble the identity of the original image more closely than other methods. Note that the second-best performing method, StyleGAN search [61], often fails completely as seen in the last row.…”

“…Vec2Face [12] uses a new bijection metric and knowledge distillation from a black-box FR model to produce realistic identity-preserving faces; however, it requires a large data set (Casia-WebFace [67]) that contains multiple images of the same identity during training. Vendrow and Vendrow [61] take advantage of a pre-trained StyleGAN [24] generator to produce face images. They perform a random search in the StyleGAN latent space to find face images with an ID vector close to the target.…”