scholarly journals Multivariate learning framework for long-term adaptation in the artificial pancreas

2018 ◽  
Vol 4 (1) ◽  
pp. 61-74 ◽  
Author(s):  
Dawei Shi ◽  
Eyal Dassau ◽  
Francis J. Doyle
Keyword(s):  
Artificial Pancreas ◽  
Learning Framework

Predicting Human Mobility with Reinforcement-Learning-Based Long-Term Periodicity Modeling

2021 ◽  
Vol 12 (6) ◽  
pp. 1-23
Author(s):  
Shuo Tao ◽  
Jingang Jiang ◽  
Defu Lian ◽  
Kai Zheng ◽  
Enhong Chen
Keyword(s):  
Reinforcement Learning ◽  
Human Mobility ◽  
Recurrent Network ◽  
Mobility Prediction ◽  
Learning Framework ◽  
Temporal Features ◽  
Wide Range ◽  
Spatio Temporal ◽  
Historical Trajectory

Mobility prediction plays an important role in a wide range of location-based applications and services. However, there are three problems in the existing literature: (1) explicit high-order interactions of spatio-temporal features are not systemically modeled; (2) most existing algorithms place attention mechanisms on top of recurrent network, so they can not allow for full parallelism and are inferior to self-attention for capturing long-range dependence; (3) most literature does not make good use of long-term historical information and do not effectively model the long-term periodicity of users. To this end, we propose MoveNet and RLMoveNet. MoveNet is a self-attention-based sequential model, predicting each user’s next destination based on her most recent visits and historical trajectory. MoveNet first introduces a cross-based learning framework for modeling feature interactions. With self-attention on both the most recent visits and historical trajectory, MoveNet can use an attention mechanism to capture the user’s long-term regularity in a more efficient way. Based on MoveNet, to model long-term periodicity more effectively, we add the reinforcement learning layer and named RLMoveNet. RLMoveNet regards the human mobility prediction as a reinforcement learning problem, using the reinforcement learning layer as the regularization part to drive the model to pay attention to the behavior with periodic actions, which can help us make the algorithm more effective. We evaluate both of them with three real-world mobility datasets. MoveNet outperforms the state-of-the-art mobility predictor by around 10% in terms of accuracy, and simultaneously achieves faster convergence and over 4x training speedup. Moreover, RLMoveNet achieves higher prediction accuracy than MoveNet, which proves that modeling periodicity explicitly from the perspective of reinforcement learning is more effective.

Development of a Convolutional Neural Network Assisted Fiber Optics Based Passive Structural Health Monitoring System

2021 ◽  
Author(s):  
Ainulla Khan ◽  
Krishnan Balasubramaniam
Keyword(s):  
Structural Health Monitoring ◽  
Fiber Optics ◽  
Health Monitoring ◽  
Alternative Energy ◽  
Power Sources ◽  
Structural Health ◽  
Learning Framework ◽  
Fbg Sensors ◽  
Non Destructive

Abstract The continuous Non-Destructive Evaluation of assets for long-term assurance of performance has led to several developments over the deployment of a Real-Time Structural Health Monitoring (SHM) system. Considering the challenges involved under the implementation of an SHM system for the applications working under harsh environmental conditions with limited access to power sources this work is aimed to contribute towards overcoming those challenges by using the noise from the structure’s machinery or any ambient source as an alternative energy source and employing Fiber Optics based sensing, for its applicability under harsh environments. The required SHM system is realized with the cross-correlation of a fully diffused noise field, sensed using the Fiber Bragg Grating (FBG) sensors at two random locations. With no control on the input received as noise, to this end, a method is developed based on a Deep Learning framework, which is aimed towards a Universal Deployment of the passive SHM system. The methodology is designed to perform the health monitoring of the system, independent of the input perturbations. The validation performed on simulation data has demonstrated the feasibility of the developed technique towards the required kind of passive SHM system.

scholarly journals Descriptive Answer Script Grading System using CNN-BiLSTM Network

2021 ◽  
Vol 9 (5) ◽  
pp. 139-144
Author(s):  
Shirien K A ◽  
Neethu George ◽  
Surekha Mariam Varghese
Keyword(s):  
Short Term Memory ◽  
Student Evaluation ◽  
Semantic Representation ◽  
Similarity Score ◽  
Test Dataset ◽  
Learning Framework ◽  
Proposed Model ◽  
Public Dataset ◽  
Long Short Term Memory

Descriptive answer script assessment and rating program is an automated framework to evaluate the answer scripts correctly. There are several classification schemes in which a piece of text is evaluated on the basis of spelling, semantics and meaning. But, lots of these aren’t successful. Some of the models available to rate the response scripts include Simple Long Short Term Memory (LSTM), Deep LSTM. In addition to that Convolution Neural Network and Bi-directional LSTM is considered here to refine the result. The model uses convolutional neural networks and bidirectional LSTM networks to learn local information of words and capture long-term dependency information of contexts on the Tensorflow and Keras deep learning framework. The embedding semantic representation of texts can be used for computing semantic similarities between pieces of texts and to grade them based on the similarity score. The experiment used methods for data optimization, such as data normalization and dropout, and tested the model on an Automated Student Evaluation Short Response Scoring, a commonly used public dataset. By comparing with the existing systems, the proposed model has achieved the state-of-the-art performance and achieves better results in the accuracy of the test dataset.

scholarly journals Classification of Micro-Calcification in Breast from Mammographic Images using Transfer Learning

2020 ◽  
Vol 8 (5) ◽  
pp. 4835-4841
Keyword(s):  
Transfer Learning ◽  
Early Identification ◽  
Breast Tumor ◽  
Early Detection Of Cancer ◽  
Medical Images ◽  
Term Survival ◽  
Learning Framework ◽  
Mammographic Images

Early detection of cancer is most important for long term survival of patient. Now a days CADx are widely used for early identification of breast cancer automatically. CAD uses significant features to identify and categorize cancer. CADx based on Convolutional Neural Network are becoming popular now a days due to extracting relevant features automatically. CNNs can be trained from scratch for medical images due to various input sizes and tumor structures. But due to limited amount of medical images available for training ,we have used transfer learning approach.We developed a deep learning framework based on CNN to discriminate the breast tumor either benign or malignant using transfer learning. We used digital mammographic images containing both views from CBIS-DDSM database. We have achived training(100%) and validation accuracy greater than 90% with minimum training and validation loss. We have also compared the reaults with transfer learning using pretrained network alexnet and googlenet on same dataset.

scholarly journals An SVM-Based Framework for Long-Term Learning Systems

2019 ◽  
Vol 33 ◽  
pp. 9915-9916
Author(s):  
Diana Benavides-Prado
Keyword(s):  
Binary Classification ◽  
Upper Bounds ◽  
Support Vector ◽  
Transfer Of Knowledge ◽  
Learning Framework ◽  
Vector Machines ◽  
Novel Method ◽  
Classification Tasks ◽  
Retention Of Knowledge

In our research, we study the problem of learning a sequence of supervised tasks. This is a long-standing challenge in machine learning. Our work relies on transfer of knowledge between hypotheses learned with Support Vector Machines. Transfer occurs in two directions: forward and backward. We have proposed to selectively transfer forward support vector coefficients from previous hypotheses as upper-bounds on support vector coefficients to be learned on a target task. We also proposed a novel method for refining existing hypotheses by transferring backward knowledge from a target hypothesis learned recently. We have improved this method through a hypothesis refinement approach that refines whilst encouraging retention of knowledge. Our contribution is represented in a long-term learning framework for binary classification tasks received sequentially one at a time.

Global Awareness Dialogue Project: Exploring Potential for Faculty Transformation Through a Professional Development Series

2019 ◽  
Vol 17 (4) ◽  
pp. 318-336
Author(s):  
Roxanna M. Senyshyn ◽  
Paula Smith
Keyword(s):  
Professional Development ◽  
Transformative Learning ◽  
Undergraduate Students ◽  
Teaching And Learning ◽  
Development Program ◽  
Global Awareness ◽  
Faculty Professional Development ◽  
Diverse Classroom ◽  
Learning Framework

This article examines a faculty professional development program aimed at engaging faculty in an ongoing discussion about global awareness and the practices surrounding teaching and learning in a linguistically and culturally diverse classroom. The program is modeled on two key components: the engagement of undergraduate students and faculty in a structured dialogue outside the classroom and the contributions of experts to faculty discussions to locate and contextualize best practices. The initiative’s potential to affect faculty perspectives is explored through a transformative learning framework. Short-term and long-term assessments show that participants were motivated to improve and implement new instructional practices.

Biohybrid Artificial Pancreas Long-Term Insulin Secretion by Devices Seeded with Canine Islets

Diabetes ◽  
1985 ◽  
Vol 34 (9) ◽  
pp. 850-854 ◽  
Author(s):  
Y. Araki ◽  
B. A. Solomon ◽  
R. M. Basile ◽  
W. L. Chick
Keyword(s):  
Insulin Secretion ◽  
Artificial Pancreas

scholarly journals A Deep Learning Framework for Automatic Meal Detection and Estimation in Artificial Pancreas Systems

Sensors ◽  
2022 ◽  
Vol 22 (2) ◽  
pp. 466
Author(s):  
John Daniels ◽  
Pau Herrero ◽  
Pantelis Georgiou
Keyword(s):  
Deep Learning ◽  
Quantile Regression ◽  
Glucose Control ◽  
Closed Loop ◽  
Artificial Pancreas ◽  
Postprandial Glucose ◽  
User Engagement ◽  
Frequent User ◽  
Learning Framework ◽  
Closed Loop Systems

Current artificial pancreas (AP) systems are hybrid closed-loop systems that require manual meal announcements to manage postprandial glucose control effectively. This poses a cognitive burden and challenge to users with T1D since this relies on frequent user engagement to maintain tight glucose control. In order to move towards fully automated closed-loop glucose control, we propose an algorithm based on a deep learning framework that performs multitask quantile regression, for both meal detection and carbohydrate estimation. Our proposed method is evaluated in silico on 10 adult subjects from the UVa/Padova simulator with a Bio-inspired Artificial Pancreas (BiAP) control algorithm over a 2 month period. Three different configurations of the AP are evaluated -BiAP without meal announcement (BiAP-NMA), BiAP with meal announcement (BiAP-MA), and BiAP with meal detection (BiAP-MD). We present results showing an improvement of BiAP-MD over BiAP-NMA, demonstrating 144.5 ± 6.8 mg/dL mean blood glucose level (−4.4 mg/dL, p< 0.01) and 77.8 ± 6.3% mean time between 70 and 180 mg/dL (+3.9%, p< 0.001). This improvement in control is realised without a significant increase in mean in hypoglycaemia (+0.1%, p= 0.4). In terms of detection of meals and snacks, the proposed method on average achieves 93% precision and 76% recall with a detection delay time of 38 ± 15 min (92% precision, 92% recall, and 37 min detection time for meals only). Furthermore, BiAP-MD handles hypoglycaemia better than BiAP-MA based on CVGA assessment with fewer control errors (10% vs. 20%). This study suggests that multitask quantile regression can improve the capability of AP systems for postprandial glucose control without increasing hypoglycaemia.

Sign in / Sign up

Export Citation Format

Share Document