Search

Your search keyword '"Szerlip, Paul"' showing total 19 results
Start Over Author "Szerlip, Paul"
19 results on '"Szerlip, Paul"'

2. Applying SVGD to Bayesian Neural Networks for Cyclical Time-Series Prediction and Inference

Author

Hu, Xinyu, Szerlip, Paul, Karaletsos, Theofanis, and Singh, Rohit

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

A regression-based BNN model is proposed to predict spatiotemporal quantities like hourly rider demand with calibrated uncertainties. The main contributions of this paper are (i) A feed-forward deterministic neural network (DetNN) architecture that predicts cyclical time series data with sensitivity to anomalous forecasting events; (ii) A Bayesian framework applying SVGD to train large neural networks for such tasks, capable of producing time series predictions as well as measures of uncertainty surrounding the predictions. Experiments show that the proposed BNN reduces average estimation error by 10% across 8 U.S. cities compared to a fine-tuned multilayer perceptron (MLP), and 4% better than the same network architecture trained without SVGD., Comment: Third workshop on Bayesian Deep Learning (NeurIPS 2018)

Published
2019

3. Pyro: Deep Universal Probabilistic Programming

Author

Bingham, Eli, Chen, Jonathan P., Jankowiak, Martin, Obermeyer, Fritz, Pradhan, Neeraj, Karaletsos, Theofanis, Singh, Rohit, Szerlip, Paul, Horsfall, Paul, and Goodman, Noah D.

Subjects
Computer Science - Machine Learning, Computer Science - Programming Languages, Statistics - Machine Learning
Abstract

Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. To scale to large datasets and high-dimensional models, Pyro uses stochastic variational inference algorithms and probability distributions built on top of PyTorch, a modern GPU-accelerated deep learning framework. To accommodate complex or model-specific algorithmic behavior, Pyro leverages Poutine, a library of composable building blocks for modifying the behavior of probabilistic programs., Comment: Submitted to JMLR MLOSS track

Published
2018

4. A Proposed Infrastructure for Adding Online Interaction to Any Evolutionary Domain

Author

Szerlip, Paul and Stanley, Kenneth O.

Subjects
Computer Science - Neural and Evolutionary Computing
Abstract

To address the difficulty of creating online collaborative evolutionary systems, this paper presents a new prototype library called Worldwide Infrastructure for Neuroevolution (WIN) and its accompanying site WIN Online (http://winark.org/). The WIN library is a collection of software packages built on top of Node.js that reduce the complexity of creating fully persistent, online, and interactive (or automated) evolutionary platforms around any domain. WIN Online is the public interface for WIN, providing an online collection of domains built with the WIN library that lets novice and expert users browse and meaningfully contribute to ongoing experiments. The long term goal of WIN is to make it trivial to connect any platform to the world, providing both a stream of online users, and archives of data and discoveries for later extension by humans or computers., Comment: Presented at WebAL-1: Workshop on Artificial Life and the Web 2014 (arXiv:1406.2507)

Published
2014

5. Unsupervised Feature Learning through Divergent Discriminative Feature Accumulation

Author

Szerlip, Paul A., Morse, Gregory, Pugh, Justin K., and Stanley, Kenneth O.

Subjects
Computer Science - Neural and Evolutionary Computing, Computer Science - Learning
Abstract

Unlike unsupervised approaches such as autoencoders that learn to reconstruct their inputs, this paper introduces an alternative approach to unsupervised feature learning called divergent discriminative feature accumulation (DDFA) that instead continually accumulates features that make novel discriminations among the training set. Thus DDFA features are inherently discriminative from the start even though they are trained without knowledge of the ultimate classification problem. Interestingly, DDFA also continues to add new features indefinitely (so it does not depend on a hidden layer size), is not based on minimizing error, and is inherently divergent instead of convergent, thereby providing a unique direction of research for unsupervised feature learning. In this paper the quality of its learned features is demonstrated on the MNIST dataset, where its performance confirms that indeed DDFA is a viable technique for learning useful features., Comment: Corrected citation formatting

Published
2014

16. Worldwide Infrastructure for Neuroevolution: A Modular Library to Turn Any Evolutionary Domain into an Online Interactive Platform

Author

Szerlip, Paul

Subjects
Neural networks; neat; hyperneat; interactive evolution; iec; novelty, Engineering, Dissertations, Academic -- Engineering and Computer Science; Engineering and Computer Science -- Dissertations, Academic
Abstract

Across many scientific disciplines, there has emerged an open opportunity to utilize the scale and reach of the Internet to collect scientific contributions from scientists and non-scientists alike. This process, called citizen science, has already shown great promise in the fields of biology and astronomy. Within the fields of artificial life (ALife) and evolutionary computation (EC) experiments in collaborative interactive evolution (CIE) have demonstrated the ability to collect thousands of experimental contributions from hundreds of users across the glob. However, such collaborative evolutionary systems can take nearly a year to build with a small team of researchers. This dissertation introduces a new developer framework enabling researchers to easily build fully persistent online collaborative experiments around almost any evolutionary domain, thereby reducing the time to create such systems to weeks for a single researcher. To add collaborative functionality to any potential domain, this framework, called Worldwide Infrastructure for Neuroevolution (WIN), exploits an important unifying principle among all evolutionary algorithms: regardless of the overall methods and parameters of the evolutionary experiment, every individual created has an explicit parent-child relationship, wherein one individual is considered the direct descendant of another. This principle alone is enough to capture and preserve the relationships and results for a wide variety of evolutionary experiments, while allowing multiple human users to meaningfully contribute. The WIN framework is first validated through two experimental domains, image evolution and a new two-dimensional virtual creature domain, Indirectly Encoded SodaRace (IESoR), that is shown to produce a visually diverse variety of ambulatory creatures. Finally, an Android application built with WIN, "filters, allows users to interactively evolve custom image effects to apply to personalized photographs, thereby introducing the first CIE application available for any mobile device. Together, these collaborative experiments and new mobile application establish a comprehensive new platform for evolutionary computation that can change how researchers design and conduct citizen science online.

Published
2015

17. Pyro: Deep Universal Probabilistic Programming.

Author

Bingham, Eli, Chen, Jonathan P., Jankowiak, Martin, Obermeyer, Fritz, Pradhan, Neeraj, Karaletsos, Theofanis, Singh, Rohit, Szerlip, Paul, Horsfall, Paul, and Goodman, Noah D.

Subjects
*PROGRAMMING languages, *PROBABILITY theory, *INFERENTIAL statistics, *STOCHASTIC analysis, *DEEP learning
Abstract

Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. To scale to large data sets and high-dimensional models, Pyro uses stochastic variational inference algorithms and probability distributions built on top of PyTorch, a modern GPU-accelerated deep learning framework. To accommodate complex or model-specific algorithmic behavior, Pyro leverages Poutine, a library of composable building blocks for modifying the behavior of probabilistic programs. [ABSTRACT FROM AUTHOR]

Published
2019

18. Association between payer status and patient-reported outcomes in adult patients with lumbar spinal stenosis treated with decompression surgery.

Author

Elsayed G, McClugage SG, Erwood MS, Davis MC, Dupépé EB, Szerlip P, Walters BC, and Hadley MN

Subjects
Adult, Aged, Back Pain surgery, Female, Humans, Male, Middle Aged, Patient Reported Outcome Measures, Quality of Life, Treatment Outcome, United States, Decompression, Surgical methods, Lumbar Vertebrae surgery, Lumbosacral Region surgery, Medicare, Spinal Stenosis surgery
Abstract

OBJECTIVE: Insurance disparities can have relevant effects on outcomes after elective lumbar spinal surgery. The aim of this study was to evaluate the association between private/public payer status and patient-reported outcomes in adult patients who underwent decompression surgery for lumbar spinal stenosis. METHODS: A sample of 100 patients who underwent surgery for lumbar spinal stenosis from 2012 to 2014 was evaluated as part of the prospectively collected Quality Outcomes Database at a single institution. Outcome measures were evaluated at 3 months and 12 months, analyzed in regard to payer status (private insurance vs Medicare/Veterans Affairs insurance), and adjusted for potential confounders. RESULTS: At baseline, patients had similar visual analog scale back and leg pain, Oswestry Disability Index, and EQ-5D scores. At 3 months postintervention, patients with government-funded insurance reported significantly worse quality of life (mean difference 0.11, p < 0.001) and more leg pain (mean difference 1.26, p = 0.05). At 12 months, patients with government-funded insurance reported significantly worse quality of life (mean difference 0.14, p < 0.001). There were no significant differences at 3 months or 12 months between groups for back pain (p = 0.14 and 0.43) or disability (p = 0.19 and 0.15). Across time points, patients in both groups showed improvement at 3 months and 12 months in all 4 functional outcomes compared with baseline (p < 0.001). CONCLUSIONS: Both private and public insurance patients had significant improvement after elective lumbar spinal surgery. Patients with public insurance had slightly less improvement in quality of life after surgery than those with private insurance but still benefited greatly from surgical intervention, particularly with respect to functional status.

Published
2018
Full Text
View/download PDF

19. Education level as a prognostic indicator at 12 months following decompression surgery for symptomatic lumbar spinal stenosis.

Author

Elsayed GA, Dupépé EB, Erwood MS, Davis MC, McClugage SG, Szerlip P, Walters BC, and Hadley MN

Subjects
Adult, Aged, Decompression, Surgical methods, Disability Evaluation, Female, Humans, Lumbosacral Region physiopathology, Lumbosacral Region surgery, Male, Middle Aged, Prospective Studies, Spinal Fusion methods, Surveys and Questionnaires, Treatment Outcome, Back Pain surgery, Lumbar Vertebrae surgery, Patient Education as Topic, Spinal Stenosis surgery
Abstract

OBJECTIVEThe goal of this study was to analyze the effect of patient education level on functional outcomes following decompression surgery for symptomatic lumbar spinal stenosis.METHODSPatients with surgically decompressed symptomatic lumbar stenosis were collected in a prospective observational registry at a single institution between 2012 and 2014. Patient education level was compared to surgical outcomes to elucidate any relationships. Outcomes were defined using the Oswestry Disability Index score, back and leg pain visual analog scale (VAS) score, and the EuroQol-5 Dimensions questionnaire score.RESULTSOf 101 patients with symptomatic lumbar spinal stenosis, 27 had no college education and 74 had a college education (i.e., 2-year, 4-year, or postgraduate degree). Preoperatively, patients with no college education had statistically significantly greater back and leg pain VAS scores when compared to patients with a college education. However, there was no statistically significant difference in quality of life or disability between those with no college education and those with a college education. Postoperatively, patients in both cohorts improved in all 4 patient-reported outcomes at 3 and 12 months after treatment for symptomatic lumbar spinal stenosis.CONCLUSIONSDespite their education level, both cohorts showed improvement in their functional outcomes at 3 and 12 months after decompression surgery for symptomatic lumbar spinal stenosis.

Published
2018
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results