155 research outputs found
Simultaneous Twin Kernel Learning Using Polynomial Transformations for Structured Prediction
No full textMany learning problems in computer vision can be posed as structured prediction problems, where the input and output instances are structured objects such as trees, graphs or strings rather than, single labels {+1, −1} or scalars. Kernel methods such as Structured Support Vector Machines , Twin Gaussian Processes (TGP), Structured Gaussian Processes, and vector-valued Reproducing Kernel Hilbert Spaces (RKHS), offer powerful ways to perform learning and inference over these domains. Positive definite kernel functions allow us to quantitatively capture similarity between a pair of instances over these arbitrary domains. A poor choice of the kernel function, which decides the RKHS feature space, often results in poor performance. Automatic kernel selection methods have been developed, but have focused only on kernels on the input domain (i.e.’one-way’). In this work, we propose a novel and efficient algorithm for learning kernel functions simultaneously, on both input and output domains. We introduce the idea of learning polynomial kernel transformations, and call this method Simultaneous Twin Kernel Learning (STKL). STKL can learn arbitrary, but continuous kernel functions, including ’one-way’ kernel learning as a special case. We formulate this problem for learning covariances kernels of Twin Gaussian Processes. Our experimental evaluation using learned kernels on synthetic and several real-world datasets demonstrate consistent improvement in performance of TGP’s.Peer reviewe
The study of open source CMSs
No full textIn this thesis, we evaluate different Open Source Content Management Systems (CMSs) and determine their appropriateness for scientific research laboratories' website content management. We describe different CMSs and evaluate them based on the following criteria: ease of installation, usability, maintenance and updates, scalability, community strength and support, user roles and workflow, security, and Web 2.0 features. We then choose one of these systems, Drupal, and demonstrate its effectiveness for two different scientific websites, Bio-1 and Vizlab. Drupal allows integrating new features using community contributed modules and easy future up-gradation. Successful implementation of both projects using Drupal highlights the importance of Open Source CMSs.M.S.Includes abstractIncludes bibliographical referencesby Chetan Gopilal Jai
System integration and image pre-processing for an automated, real-time identification and monitoring system for coral reef fish
No full textIn this work we build an underwater vision system capable of monitoring the activities of fish found near coral reefs. We propose a unique hardware platform capable of monitoring a volume of water in a very efficient and cost effective way. We also develop algorithms required to take advantage of such a system. There are three main contributions of this work, which are; (1) using two right-angled camera’s to capture underwater image sequences, (2) developing algorithms to track and pre-process images for recognition (3) and demonstrating that we can recognize fish families or in some cases exact fish species using fish shape(with size, color and pattern features to be added later). We conclude from this work that using just two cameras in a right-angled setup is a cheap and effective way of monitoring fish activities in general. It is cost effective when compared to using multiple cameras and also less computationally intensive. We developed and modified our approach based on observations we made while testing this setup and accommodated these modifications in our software. We installed this system at the artificial coral reef in the New York Aquarium and periodically collected image sequences for processing. We demonstrate our results on the collected sequences and show pre-processing results on them. We also demonstrate, using shape feature from a fish sequence we collected at the aquarium (using cross-validation); that we can recognize fish families or in some cases exact species using those features.M.S.Includes bibliographical referencesby Chetan Tond
Improving Bicycle Safety through Automated Real-Time Vehicle Detection
No full textThe manner in which people use bicycles has changed very little since their invention in 1817. In that time, though, roadways have become congested with a dramatically less environmentally friendly mode of transportation: automobiles. These vehicles and the motorists who drive them represent, at times, a serious threat to the safety of both road cycling enthusiasts and bicycle commuters alike. Since bikers typically ride with the ow of trac, the most dangerous situation for them is when they are being passed by a motorist from behind. As a result, a biker must spend a substantial amount of her cognitive and physical ability to periodically scan for rear-approaching vehicles, reducing her capacity to handle the bicycle safely and maintain continual awareness for both the forward and rearward situations.To improve road cycling safety, we present a system that augments a standard bicycle with audio and video sensing, and computational capabilities. This Cyber-Physical bicycle system continuously senses the environment behind a biker, processes the sensed data utilizing audio processing and computer vision techniques, automatically detects the occurrences of rear-approaching vehicles, and alerts the biker in real time prior to the encounter. In this paper we present (i) the design of our prototype Cyber-Physical bicycle system and (ii) the results of our evaluation using video and audio traces collected from bikers. These results demonstrate both the feasibility of the system, exhibiting a high degree of detection accuracy while operating under the real-time and energy constraints of the problem scenario.Technical report DCS-TR-66
Integrated Continuous Biomanufacturing: Perspectives from the Manufacturing Floor
No full textPlease click Additional Files below to see the full abstrac
Computer programs for modeling mammalian cell batch and fed-batch cultures using logistic equations
No full text
Special Section on Continuous Bioprocessing and Welcome Professor Liu to the Editorial Board
No full textBioprocess optimization for recombinant protein production from mammalian cells
No full textMammalian cells are being increasingly used to manufacture complex therapeutic proteins given their ability to properly fold and glycosylate these proteins. However, protein yields are low and process enhancements are necessary to ensure economically viable processes. Methods for yield improvement, bioprocess development acceleration and rapid quantification and monitoring of cell metabolism were investigated in this study. Recognizing the adverse effect of high PCO₂ on cell growth, metabolism and protein productivity, a novel PCO₂ reduction strategy based on NaHCO₃ elimination was investigated that decreased PCO₂ by 65-72%. This was accompanied by 68-123% increases in growth rate and 58-92% increases in productivity. To enable rapid and robust data analysis from early stage process development experiments, logistic equations were used to effectively describe the kinetics of batch and fed-batch cultures. Substantially improved specific rate estimates were obtained from the logistic equations when compared with current modeling approaches. Metabolic flux analysis was used to obtain quantitative information on cellular metabolism and the validity of using the balancing method for flux estimation was verified with data from isotope tracer studies. Error propagation from prime variables into specific rates and metabolic fluxes was quantified using Monte-Carlo analysis which indicated 8-22% specific rate error for 5-15% error in prime variable measurement. While errors in greater metabolic fluxes were similar to those in the corresponding specific rates, errors in the lesser metabolic fluxes were extremely sensitive to greater specific rate errors such that lesser fluxes were no longer representative of cellular metabolism. The specific rate to metabolic flux error relationship could be accurately described by the corresponding normalized sensitivity coefficient. A framework for quasi-real-time estimation of metabolic fluxes was proposed and implemented to serve as a bioprocess monitoring and early warning system. Methods for real-time oxygen uptake and carbon dioxide production rate estimation were developed that enabled rapid flux estimation. This framework was used to characterize cellular response to pH and dissolved oxygen changes in a process development experiment and can readily be applied to a manufacturing bioreactor. Overall, the approaches for protein productivity enhancement and rapid metabolism monitoring developed in this study are general with potential for widespread application to laboratory and manufacturing-scale perfusion and fed-batch mammalian cell cultivations.Applied Science, Faculty ofChemical and Biological Engineering, Department ofGraduat
Accurate kinetic parameter estimation during progress curve analysis of systems with endogenous substrate production
No full text- …
