ISLAB/CAISR - User contributions [en]

FLBench: A Comprehensive Experimental Evaluation of Federated Learning Frameworks

2022-10-03T09:07:24Z

Jens: Created page with "{{StudentProjectTemplate |Summary=Exploring Federated Learning Frameworks |Keywords=Federated Learning |Prerequisites=- DevOps: Good installation and configuration.    ..."

{{StudentProjectTemplate
|Summary=Exploring Federated Learning Frameworks
|Keywords=Federated Learning
|Prerequisites=- DevOps: Good installation and configuration.   

- Programming: Pytorch, and Python. 
|Supervisor=Sadi Alawadi, Jens Lundström
|Level=Master
|Status=Open
}}
The advent of distributed Machine Learning (ML) promoted sophisticated analytics at the network's edge. This decentralized and large-scale ML architecture is known as Federated Learning (FL). FL aims to enable multiple actors to build a common and robust ML model over multiple local dataset. Furthermore, the new wave of FL frameworks promotes data privacy, security, access rights, and access to heterogeneous data. However, the variety of these frameworks require an experimental evaluation of performance analysis. Therefore, this project aims to analyze, evaluate, compare and conclude the popular federated learning frameworks extensively (A similar comparison paper can be found here: https://link.springer.com/content/pdf/10.1007/s10586-021-03240-4.pdf ). The main intended tasks of this project are: 

Evaluating the following open-source federated learning frameworks.

Paddle Federated Learning Framework (https://github.com/PaddlePaddle/PaddleFL) 

PySyft Framework /pygrid (https://github.com/OpenMined/PySyft  ) 

Flower (https://github.com/adap/flower)

TensorFlow FL (https://github.com/tensorflow/federated )

FEDn (https://github.com/scaleoutsystems/fedn )

Intel FL (https://github.com/intel/openfl)

FATE (https://github.com/FederatedAI/FATE  )

Benchmarking suit (Experiment design), i.e., network architecture (the ML model, e.g., LSTM, CNN, etc.), the used datasets, benchmark tool/framework. (Mnist, Cifer10 & 100, IMDB, for IoT data CASA activity recognition).  

Theoretically comparing the federated algorithm they support (FedAVG, FedProx, etc.), cross-device and cross-silo, horizontal and vertical federated learning. Also, open-source, Diversified Computing Paradigms (Standalone simulation, Distributed computing, on-device training), ML heterogeneity (Pytorch, TensorFlow, MXnet,...etc.), development coding language, Framework's timeline.  

Use of existing and development of comparison criteria, performance (task per time, aka, throughput), resources consumption (CPU, Memory, GPU), convergence, deployment effort, Flexibility, accuracy, scalability. 

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T09:02:17Z

Jens:

{{StudentProjectTemplate
|Summary=Time-series GAN and generation of synthetic electrical health records
|Keywords=GAN, Synthetic data, Time-series
|References=Papers: https://ieeexplore.ieee.org/abstract/document/8975823 https://proceedings.neurips.cc/paper/2019/file/c9efe5f26cd17ba6216bbe2a7d26d490-Paper.pdf https://arxiv.org/pdf/2009.09283.pdf https://dl.acm.org/doi/abs/10.1145/2810103.2813687

Data: https://lcp.mit.edu/mimic
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T09:01:43Z

Jens:

{{StudentProjectTemplate
|Summary=Time-series GAN and generation of synthetic electrical health records
|References=Papers: https://ieeexplore.ieee.org/abstract/document/8975823 https://proceedings.neurips.cc/paper/2019/file/c9efe5f26cd17ba6216bbe2a7d26d490-Paper.pdf https://arxiv.org/pdf/2009.09283.pdf https://dl.acm.org/doi/abs/10.1145/2810103.2813687

Data: https://lcp.mit.edu/mimic
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T08:57:44Z

Jens:

{{StudentProjectTemplate
|Summary=Time-series GAN and generation of synthetic electrical health records
|References=PPGAN: Privacy-preserving generative adversarial networks: PPGAN: Privacy-Preserving Generative Adversarial Network

Time-series Generative Adversarial Network.

About MIMIC | MIMIC (mit.edu)

Subverting Privacy-Preserving GANs: Hiding Secrets in Sanitized Images

Privacy-Preserving Deep Learning.
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T08:57:11Z

Jens:

{{StudentProjectTemplate
|Summary=Time-series GAN and generation of synthetic electrical health records
|References=[1] PPGAN: Privacy-preserving generative adversarial networks: PPGAN: Privacy-Preserving Generative Adversarial Network

[2] Time-series Generative Adversarial Network.

[3] About MIMIC | MIMIC (mit.edu)

[4] Subverting Privacy-Preserving GANs: Hiding Secrets in Sanitized Images

[5] Privacy-Preserving Deep Learning.
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T08:55:52Z

Jens:

{{StudentProjectTemplate
|Summary=Time-series GAN and generation of synthetic electrical health records
|References=1. PPGAN: Privacy-preserving generative adversarial networks: PPGAN: Privacy-Preserving Generative Adversarial Network | IEEE Conference Publication | IEEE Xplore

2. Time-series Generative Adversarial Network: https://proceedings.neurips.cc/paper/2019/file/c9efe5f26cd17ba6216bbe2a7d26d490-Paper.pdf 3. About MIMIC | MIMIC (mit.edu)

4. Subverting Privacy-Preserving GANs: Hiding Secrets in Sanitized Images: https://arxiv.org/pdf/2009.09283.pdf 5. Privacy-Preserving Deep Learning: https://dl.acm.org/doi/abs/10.1145/2810103.2813687
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Privacy-Preserved Generator for Generating Synthetic EHR data

2022-10-03T08:46:38Z

Jens: Created page with "{{StudentProjectTemplate |Summary=Time-series GANs are applicable for generating synthetic electrical health records |References=1. PPGAN: Privacy-preserving generative advers..."

{{StudentProjectTemplate
|Summary=Time-series GANs are applicable for generating synthetic electrical health records
|References=1. PPGAN: Privacy-preserving generative adversarial networks: PPGAN: Privacy-Preserving Generative Adversarial Network | IEEE Conference Publication | IEEE Xplore

2. Time-series Generative Adversarial Network: https://proceedings.neurips.cc/paper/2019/file/c9efe5f26cd17ba6216bbe2a7d26d490-Paper.pdf 3. About MIMIC | MIMIC (mit.edu)

4. Subverting Privacy-Preserving GANs: Hiding Secrets in Sanitized Images: https://arxiv.org/pdf/2009.09283.pdf 5. Privacy-Preserving Deep Learning: https://dl.acm.org/doi/abs/10.1145/2810103.2813687
|Supervisor=Atiye Sadat Hashemi, Jens Lundström, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Access to high-quality big datasets for improving deep learning (DL)-based models is a big challenge, more specifically in overly sensitive applications such as healthcare systems where maintaining data privacy is a necessity. Synthetic data generation is a principal tool for various users from researchers who leverage data for models’ training, to educators who aim to teach statistical approaches. The aim of using synthetic data can be categorized into several essential groups such as protecting privacy. However, for generating synthetic data using DL models (like generative adversarial networks (GANs)) we need the training data, and it has been proved that the gradient parameters of these models can remember the training data. For focusing on the privacy-preserving issue of training data in synthetic health data generation, we are going to modify the idea of privacy-preserved GANs [1] to suitable GANs for time series data [2]. Time-series GANs are applicable for generating synthetic electrical health records (EHRs) [3]. In this master thesis, we aim to study different differential privacy-preserving methods to add well-designed noise to the gradients during the training phase of time-series GANs.

Conclusion: In this master thesis, we aim to study privacy-preserving approaches in deep learning and develop a model that preserves the privacy of training data in the processes of generating synthetic data. The title of this thesis in detail is ‘the privacy-preserving aspect of generating synthetic electrical health records (Synthetic-EHRs) using time-series generative adversarial networks (Time-GANs).

Federated Learning Aggregation Strategies by Weight Exploration

2022-10-03T08:42:52Z

Jens:

{{StudentProjectTemplate
|Summary=Investigation of aggregation strategies for federated learning
|Keywords=Federated Learning, Aggregation Strategies, Decentralized Learning
|References=Li, Tian, et al. "Federated learning: Challenges, methods, and future directions." IEEE Signal Processing Magazine 37.3 (2020): 50-60.
Kairouz, Peter, et al. "Advances and open problems in federated learning." Foundations and Trends® in Machine Learning 14.1–2 (2021): 1-210.
Rieke, Nicola, et al. "The future of digital health with federated learning." NPJ digital medicine 3.1 (2020): 1-7.
Li, Tian, et al. "Federated optimization in heterogeneous networks." Proceedings of Machine Learning and Systems 2 (2020): 429-450.
|Supervisor=Jens Lundström, Amira Soliman, Sadi Alawadi
|Examiner=Slawomir Nowaczyk
|Level=Master
|Status=Open
}}
Since the advent of Federated Learning during recent years, industries, institutions and the research community are able to train multiple models on data that stays locally and then aggregate model parameters to form a global model. This set of methods has several advantages including the ability to reduce the burden on a single server instance as well as to preserve privacy which could be valuable when working with highly sensitive data such as Electronic Health Records (EHR). There are several strategies proposed to perform the federated optimization where the most simple is when the server aggregates parameters (e.g. neural network weights) by computing the average of parameters. Past research has shown that despite the effectiveness the simpler federated optimization strategies are sensitive to the computational resource available at each local node and the statistical heterogeneity of the data (since most machine learning methods rely on the data to be identically distributed). Therefore many extensions to the simpler federated optimization have been proposed to measure and communicate the local heterogeneity, which could for some applications breach data privacy. This master thesis is about to study, understand, test and develop innovative and alternative methods based on implicit ways of using the statistical heterogeneity from each local node.

Explainable AI by Training Introspection

2022-10-03T08:41:40Z

Jens: Created page with "{{StudentProjectTemplate |Summary=Research and development of novel XAI methods based on training process information |Keywords=XAI, Neural Networks |Supervisor=Jens Lundströ..."

{{StudentProjectTemplate
|Summary=Research and development of novel XAI methods based on training process information
|Keywords=XAI, Neural Networks
|Supervisor=Jens Lundström, Peyman Mashhadi, Amira Soliman, Atiye Sadat Hashemi
|Level=Master
|Status=Open
}}
As machine learning has become increasingly successful in commercial applications during the last decades, the demand for model explainability and interpretability also emerge. In many occasions, for a decision support system to be credible and useful the predicted decision support needs to follow with explainability. This need has sparked enormous activity in the field of Explainable AI (XAI) both for the industry and in AI/ML research for a couple of years. The focus of current XAI methods aims at utilizing the end result of the training process, i.e. the final trained model. In the master thesis we explore the hypothesized potential of XAI to be revealed by exploring the full trajectory of the model training process. The thesis will explore different data modalities, types of models and explainability aspects.

Quantifying exercise-induced muscle fatigue by machine learning

2022-10-03T08:35:51Z

Jens: Created page with "{{StudentProjectTemplate |Summary=Exploring machine learning methods on an EMG muscle fatigue pipeline |Keywords=EMG, Machine Learning, Fatigue |References=Yousif, Hayder A., ..."

{{StudentProjectTemplate
|Summary=Exploring machine learning methods on an EMG muscle fatigue pipeline
|Keywords=EMG, Machine Learning, Fatigue
|References=Yousif, Hayder A., et al. "Assessment of muscles fatigue based on surface EMG signals using machine learning and statistical approaches: a review." IOP conference series: materials science and engineering. Vol. 705. No. 1. IOP Publishing, 2019.

Karlik, Bekir. "Machine learning algorithms for characterization of EMG signals." International Journal of Information and Electronics Engineering 4.3 (2014): 189.

Rampichini, S., Vieira, T. M., Castiglioni, P., & Merati, G. (2020). Complexity analysis of surface electromyography for assessing the myoelectric manifestation of muscle fatigue: A review. Entropy, 22(5), 529.

Carroll, T. J., Taylor, J. L., & Gandevia, S. C. (2017). Recovery of central and peripheral neuromuscular fatigue after exercise. Journal of Applied Physiology, 122(5), 1068-1076.

Yousefi, J., & Hamilton-Wright, A. (2014). Characterizing EMG data using machine-learning tools. Computers in biology and medicine, 51, 1-13.
|Supervisor=Jens Lundström
|Level=Master
|Status=Open
}}
Electromyography (EMG), the measurement of the electrical activity in the muscles as a response to nerve stimulation is an advanced and prominent tool for assessing muscle fatigue. During sports it is important to assess exactly when and to which extent muscle fatigue occurs in order to optimize recovery, effort distribution, and planning of the training process. Relative to manual rules machine learning has been shown to be highly effective to characterize and detect muscle fatigue, yet a few challenges remain despite the success of surface-based EMG. Variability between subjects and sensor placements pose signals complexity and analysis could potentially benefit from the more recent and advanced machine learning methods such as advanced time-series models and the concept of attention. This project is about surveying and exploring state-of-the-art methods and systematically, theoretically, and practically test the applicability and performance of more recent machine learning methods on an existing EMG to muscle fatigue pipeline. Also, the use of XAI for the characterization of the signals is desired to develop the interpretation of the developed models. You will be working with a local startup sports-tech company and the work also included to do hands-on experiments with sensors and embedded systems.

Federated Learning Aggregation Strategies by Weight Exploration

2022-10-03T08:31:51Z

Jens:

{{StudentProjectTemplate
|Summary=Investigation of aggregation strategies for federated learning
|Keywords=Federated Learning, Aggregation Strategies, Decentralized Learning
|References=Li, Tian, et al. "Federated learning: Challenges, methods, and future directions." IEEE Signal Processing Magazine 37.3 (2020): 50-60.
Kairouz, Peter, et al. "Advances and open problems in federated learning." Foundations and Trends® in Machine Learning 14.1–2 (2021): 1-210.
Rieke, Nicola, et al. "The future of digital health with federated learning." NPJ digital medicine 3.1 (2020): 1-7.
Li, Tian, et al. "Federated optimization in heterogeneous networks." Proceedings of Machine Learning and Systems 2 (2020): 429-450.
|Supervisor=Jens Lundström, Amira Soliman, and Sadi Alawadi
|Examiner=Slawomir Nowaczyk
|Level=Master
|Status=Open
}}
Since the advent of Federated Learning during recent years, industries, institutions and the research community are able to train multiple models on data that stays locally and then aggregate model parameters to form a global model. This set of methods has several advantages including the ability to reduce the burden on a single server instance as well as to preserve privacy which could be valuable when working with highly sensitive data such as Electronic Health Records (EHR). There are several strategies proposed to perform the federated optimization where the most simple is when the server aggregates parameters (e.g. neural network weights) by computing the average of parameters. Past research has shown that despite the effectiveness the simpler federated optimization strategies are sensitive to the computational resource available at each local node and the statistical heterogeneity of the data (since most machine learning methods rely on the data to be identically distributed). Therefore many extensions to the simpler federated optimization have been proposed to measure and communicate the local heterogeneity, which could for some applications breach data privacy. This master thesis is about to study, understand, test and develop innovative and alternative methods based on implicit ways of using the statistical heterogeneity from each local node.

Federated Learning Aggregation Strategies by Weight Exploration

2022-10-03T08:29:29Z

Jens:

{{StudentProjectTemplate
|Summary=Investigation of aggregation strategies for federated learning whilst maintaining data privacy.
|Keywords=Federated Learning, Aggregation Strategies, Decentralized Learning
|References=Li, Tian, et al. "Federated learning: Challenges, methods, and future directions." IEEE Signal Processing Magazine 37.3 (2020): 50-60.
Kairouz, Peter, et al. "Advances and open problems in federated learning." Foundations and Trends® in Machine Learning 14.1–2 (2021): 1-210.
Rieke, Nicola, et al. "The future of digital health with federated learning." NPJ digital medicine 3.1 (2020): 1-7.
Li, Tian, et al. "Federated optimization in heterogeneous networks." Proceedings of Machine Learning and Systems 2 (2020): 429-450.
|Supervisor=Jens Lundström, Amira Soliman, and Sadi Alawadi
|Examiner=Slawomir Nowaczyk
|Level=Master
|Status=Open
}}
Since the advent of Federated Learning during recent years, industries, institutions and the research community are able to train multiple models on data that stays locally and then aggregate model parameters to form a global model. This set of methods has several advantages including the ability to reduce the burden on a single server instance as well as to preserve privacy which could be valuable when working with highly sensitive data such as Electronic Health Records (EHR). There are several strategies proposed to perform the federated optimization where the most simple is when the server aggregates parameters (e.g. neural network weights) by computing the average of parameters. Past research has shown that despite the effectiveness the simpler federated optimization strategies are sensitive to the computational resource available at each local node and the statistical heterogeneity of the data (since most machine learning methods rely on the data to be identically distributed). Therefore many extensions to the simpler federated optimization have been proposed to measure and communicate the local heterogeneity, which could for some applications breach data privacy. This master thesis is about to study, understand, test and develop innovative and alternative methods based on implicit ways of using the statistical heterogeneity from each local node.

Taha Khan

2018-01-03T13:12:51Z

Jens:

{{Person
|Family Name=Khan
|Given Name=Taha
|Title=PhD
|Cell Phone=+46 729 773 807
|Position=Posdoctoral Research Fellow
|Email=taha.khan@hh.se
|Image=Head-unknown.jpg
|Country=Sweden
|Office=F504
|Affiliation=Halmstad University, Department of Intelligent Systems Laboratory
}}
{{AssignProjects
|project=HMC2
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignApplicationAreas
|ApplicationArea=Health Technology
}}
{{AssignExaminer
|Examiner=Artificial Intelligence (7.5 credits)
}}
{{AssignTeacher
|Teacher=Linux System Administration (7.5 credits)
}}
{{AssignTeacher
|Teacher=Artificial Intelligence (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Taha Khan

2018-01-03T13:12:10Z

Jens: Created page with "{{Person |Family Name=Khan |Given Name=Taha |Title=PhD |Cell Phone=+46 729 773 807 |Position=Posdoctoral Research Fellow |Email=taha.khan@hh.se |Country=Sweden |Office=F504 |A..."

{{Person
|Family Name=Khan
|Given Name=Taha
|Title=PhD
|Cell Phone=+46 729 773 807
|Position=Posdoctoral Research Fellow
|Email=taha.khan@hh.se
|Country=Sweden
|Office=F504
|Affiliation=Halmstad University, Department of Intelligent Systems Laboratory
}}
{{AssignProjects
|project=HMC2
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignApplicationAreas
|ApplicationArea=Health Technology
}}
{{AssignExaminer
|Examiner=Artificial Intelligence (7.5 credits)
}}
{{AssignTeacher
|Teacher=Linux System Administration (7.5 credits)
}}
{{AssignTeacher
|Teacher=Artificial Intelligence (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Rebeen Hamad

2018-01-02T14:52:11Z

Jens:

{{Person
|Family Name=Hamad
|Given Name=Rebeen
|Title=M.Sc.
|Phone=+46 35 167128
|Cell Phone=+46 766 08 3438
|Position=Licentiate degree student
|Email=rebeen.ali_hamad@hh.se
|Image=Head-unknown.jpg
|Country=Sweden
|Office=E504
|Affiliation=Halmstad University
|Revision Date=2018-01-02
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignApplicationAreas
|ApplicationArea=Smart Environments
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Rebeen Hamad

2018-01-02T14:51:19Z

Jens:

{{Person
|Family Name=Hamad
|Given Name=Rebeen
|Title=M.Sc
|Phone=+46 35 167128
|Cell Phone=+46 766 08 3438
|Position=Licentiate degree student
|Email=rebeen.ali_hamad@hh.se
|Image=Head-unknown.jpg
|Country=Sweden
|Office=E504
|Affiliation=Halmstad University
|Revision Date=2018-01-02
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignApplicationAreas
|ApplicationArea=Smart Environments
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Rebeen Hamad

2018-01-02T14:50:49Z

Jens:

{{Person
|Family Name=Hamad
|Given Name=Rebeen
|Title=Ms.SC.
|Phone=+46 35 167128
|Cell Phone=+46 766 08 3438
|Position=Licentiate degree student
|Email=rebeen.ali_hamad@hh.se
|Image=Head-unknown.jpg
|Country=Sweden
|Office=E504
|Affiliation=Halmstad University
|Revision Date=2018-01-02
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignApplicationAreas
|ApplicationArea=Smart Environments
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Rebeen Hamad

2018-01-02T14:50:21Z

Jens:

{{Person
|Family Name=Hamad
|Given Name=Rebeen
|Title=MsC
|Phone=+46 35 167128
|Cell Phone=+46 766 08 3438
|Position=Licentiate degree student
|Email=rebeen.ali_hamad@hh.se
|Image=Head-unknown.jpg
|Country=Sweden
|Office=E504
|Affiliation=Halmstad University
|Revision Date=2018-01-02
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignApplicationAreas
|ApplicationArea=Smart Environments
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Rebeen Hamad

2018-01-02T14:49:14Z

Jens: Created page with "{{Person |Family Name=Hamad |Given Name=Rebeen |Title=MsC |Phone=+46 35 167128 |Cell Phone=+46 766 08 3438 |Position=Licentiate degree student |Email=rebeen.ali_hamad@hh.se |C..."

{{Person
|Family Name=Hamad
|Given Name=Rebeen
|Title=MsC
|Phone=+46 35 167128
|Cell Phone=+46 766 08 3438
|Position=Licentiate degree student
|Email=rebeen.ali_hamad@hh.se
|Country=Sweden
|Office=E504
|Affiliation=Halmstad University
|Revision Date=2018-01-02
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignApplicationAreas
|ApplicationArea=Smart Environments
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertProjects}}

Karl Iagnemma

2018-01-02T13:17:47Z

Jens:

{{Person
|Family Name=Iagnemma
|Given Name=Karl
|Title=PhD
|Position=Guest Professor
|Image=karl.jpg
|Affiliation=nuTonomy
|url=http://web.mit.edu/mobility/people/karl.html
|Subject=
}}
[[Category:alumni]]


{{ShowPerson}}

Karl Iagnemma

2018-01-02T13:12:01Z

Jens:

{{Person
|Family Name=Iagnemma
|Given Name=Karl
|Title=PhD
|Position=Guest Professor
|Image=karl.jpg
|url=http://web.mit.edu/mobility/people/karl.html
|Subject=
}}
[[Category:alumni]]


{{ShowPerson}}

Iulian Carpatorea

2018-01-02T13:11:34Z

Jens:

{{Person
|Family Name=Carpatorea
|Given Name=Iulian
|Title=M.Sc.
|Phone=+4635167926
|Position=PhD. Candidate
|Email=iulian.carpatorea@hh.se
|Image=Iulian_Carpatorea_web.jpg
|Office=E513
|Subject=Computer Science
}}
{{AssignProjects
|project=fuelFEET
}}
{{AssignSubjectAreas
|SubjectArea=Computer Science
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignSubjectAreas
|SubjectArea=Signal Analysis
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}

__NOTOC__
{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}

[[Category:alumni]]

Kenneth Nilsson

2018-01-02T13:11:11Z

Jens:

{{Person
|Position=Associate Professor
|Image=Kenneth head medium.jpg
|Given Name=Kenneth
|Family Name=Nilsson
|Subject=Digital signal processing
|Title=PhD
|Office=E518
|Phone=+46 35 167136
|faxnr=+46 35 120348
|Email=kenneth.nilsson@hh.se
}}
{{AssignSubjectAreas
|SubjectArea=Signal Analysis
}}
{{AssignProjects}}
{{AssignExaminer
|Examiner=Degree Project (15 credits)
}}
{{AssignExaminer
|Examiner=Bachelor Thesis (15 credits)
}}

{{ShowPerson}}
{{InsertTeacher}}
{{InsertSubjAreas}}

{{PublicationsList}}
[[Category:alumni]]

Rune Prytz

2018-01-02T13:09:56Z

Jens:

{{Person
|Family Name=Prytz
|Given Name=Rune
|Title=M.Sc.
|Position=Industrial PhD Student
|Email=rune.prytz@volvo.com
|Image=Rune Prytz web.jpg
|Office=E508
|Affiliation=Volvo AB
|Subject=Vehicle Diagnostics
}}
{{AssignProjects
|project=ReDi2Service
}}
{{AssignProjects
|project=InnoMerge
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}

__NOTOC__
{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}
[[Category:alumni]]

Comprehending low-dimensional manifolds of temporal data from the home

2017-10-05T09:17:04Z

Jens:

{{StudentProjectTemplate
|Summary=Study and development of tools and methods for the visualization of (temporal) human activity patterns.
|Keywords=Visualization, Dimensionality Reduction, Manifold learning
|References=Maaten, L. V. D., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research, 9(Nov), 2579-2605.

Lundström, J., Järpe, E., & Verikas, A. (2016). Detecting and exploring deviating behaviour of smart home residents. Expert Systems with Applications, 55, 429-440.

Rauber, P. E., Falcão, A. X., & Telea, A. C. (2016). Visualizing time-dependent data using dynamic t-SNE. Proc. EuroVis Short Papers, 2(5).

Cheng, J., Liu, H., Wang, F., Li, H., & Zhu, C. (2015). Silhouette analysis for human action recognition based on supervised temporal t-sne and incremental learning. IEEE Transactions on Image Processing, 24(10), 3203-3217.
|Prerequisites=Completed courses in basic machine learning is required.
|Supervisor=Jens Lundström, Eric Järpe, Rebeen Hamad
|Level=Master
|Status=Open
}}
Data generated from sensors deployed in an in-home setting can be processed to characterize human activity patterns. These patterns can further be used for smart home services such as fall detection, task reminders and messages for the motivation of exercise. Despite the direct use of activity patterns there are few tools and methods available for the visual interpretation, especially the temporal variations in high-dimensional human activity patterns lack proper visualization.

Researchers in the area of Smart Homes (at the ISDD department) study and develop methods for modeling human activity patterns and now calls for student(s) to perform master thesis work in the area of model comprehensibility by studying methods for visualization. As a student in this project you are expected to be working by four suggested work packages:

1. Background study on visualization of human activity patterns and related research.
2. Practical tests on visualization methods on medium size datasets.
3. Investigation (test, development and validation) on how spatio-temporal components can be (better) visualized on real-life datasets.
The result is expected to include investigation results and conclusions on how high-dimensional human activity patterns could be visualized for better interpretation.

Comprehending low-dimensional manifolds of temporal data from the home

2017-10-05T09:14:20Z

Jens: Created page with "{{StudentProjectTemplate |Summary=Comprehending low-dimensional manifolds of temporal data from the home |Keywords=Visualization, Dimensionality Reduction, Manifold learning |..."

{{StudentProjectTemplate
|Summary=Comprehending low-dimensional manifolds of temporal data from the home
|Keywords=Visualization, Dimensionality Reduction, Manifold learning
|TimeFrame=Winter2017 / Spring 2018
|References=Maaten, L. V. D., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research, 9(Nov), 2579-2605.

Lundström, J., Järpe, E., & Verikas, A. (2016). Detecting and exploring deviating behaviour of smart home residents. Expert Systems with Applications, 55, 429-440.

Rauber, P. E., Falcão, A. X., & Telea, A. C. (2016). Visualizing time-dependent data using dynamic t-SNE. Proc. EuroVis Short Papers, 2(5).

Cheng, J., Liu, H., Wang, F., Li, H., & Zhu, C. (2015). Silhouette analysis for human action recognition based on supervised temporal t-sne and incremental learning. IEEE Transactions on Image Processing, 24(10), 3203-3217.
|Prerequisites=Completed courses in basic machine learning is required.
|Supervisor=Jens Lundström, Eric Järpe, Rebeen Hamad
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
Data generated from sensors deployed in an in-home setting can be processed to characterize human activity patterns. These patterns can further be used for smart home services such as fall detection, task reminders and messages for the motivation of exercise. Despite the direct use of activity patterns there are few tools and methods available for the visual interpretation, especially the temporal variations in high-dimensional human activity patterns lack proper visualization.

Researchers in the area of Smart Homes (at the ISDD department) study and develop methods for modeling human activity patterns and now calls for student(s) to perform master thesis work in the area of model comprehensibility by studying methods for visualization. As a student in this project you are expected to be working by four suggested work packages:

1. Background study on visualization of human activity patterns and related research.
2. Practical tests on visualization methods on medium size datasets.
3. Investigation (test, development and validation) on how spatio-temporal components can be (better) visualized on real-life datasets.
The result is expected to include investigation results and conclusions on how high-dimensional human activity patterns could be visualized for better interpretation.

Jens Lundström

2016-10-17T07:40:37Z

Jens:

{{Person
|Family Name=Lundström
|Given Name=Jens
|Title=PhD
|Phone=+46 72 396 00 23
|Position=Assistant Senior Lecturer
|Email=jens.lundstrom@hh.se
|Image=jens.png
|Country=Sweden
|Office=E509
|Affiliation=ISLAB, CAISR
|Revision Date=2016-10-17
|Subject=Information Technology
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignProjects
|project=PPQ
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignApplicationAreas
|ApplicationArea=Health Technology
}}
{{AssignExaminer
|Examiner=TV4002 Perspektiv på hälsoinnovation
}}
{{AssignExaminer
|Examiner=TV6001 Utvecklingsprojekt med fokus på hälsoinnovation
}}
{{AssignTeacher
|Teacher=Design of Embedded and Intelligent Systems (15 credits)
}}
{{AssignTeacher
|Teacher=Sensor System (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}

Jens Lundström

2016-10-17T07:37:29Z

Jens:

{{Person
|Family Name=Lundström
|Given Name=Jens
|Title=PhD
|Phone=+46 72 396 00 23
|Position=Assistant Senior Lecturer
|Email=jens.lundstrom@hh.se
|Image=jens.png
|Country=Sweden
|Office=E509
|Affiliation=ISLAB, CAISR
|Revision Date=2016-10-17
|Subject=Information Technology
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignProjects
|project=PPQ
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignSubjectAreas
|SubjectArea=Smart Environments
}}
{{AssignSubjectAreas
|SubjectArea=Health Technology
}}
{{AssignExaminer
|Examiner=TV4002 Perspektiv på hälsoinnovation
}}
{{AssignExaminer
|Examiner=TV6001 Utvecklingsprojekt med fokus på hälsoinnovation
}}
{{AssignTeacher
|Teacher=Design of Embedded and Intelligent Systems (15 credits)
}}
{{AssignTeacher
|Teacher=Sensor System (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}

File:Jens.png

2016-10-17T07:36:48Z

Jens: Jens Lundström

Jens Lundström

2016-10-17T07:35:48Z

Jens:

{{Person
|Family Name=Lundström
|Given Name=Jens
|Title=PhD
|Phone=+46 72 396 00 23
|Position=Assistant Senior Lecturer
|Email=jens.lundstrom@hh.se
|Image=jenspicture.jpeg
|Country=Sweden
|Office=E509
|Affiliation=ISLAB, CAISR
|Revision Date=2016-10-17
|Subject=Information Technology
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignProjects
|project=PPQ
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignSubjectAreas
|SubjectArea=Smart Environments
}}
{{AssignSubjectAreas
|SubjectArea=Health Technology
}}
{{AssignExaminer
|Examiner=TV4002 Perspektiv på hälsoinnovation
}}
{{AssignExaminer
|Examiner=TV6001 Utvecklingsprojekt med fokus på hälsoinnovation
}}
{{AssignTeacher
|Teacher=Design of Embedded and Intelligent Systems (15 credits)
}}
{{AssignTeacher
|Teacher=Sensor System (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}

Jens Lundström

2016-10-17T07:34:59Z

Jens:

{{Person
|Family Name=Lundström
|Given Name=Jens
|Title=PhD
|Phone=+46 72 396 00 23
|Position=Assistant Senior Lecturer
|Email=jens.lundstrom@hh.se
|Image=jenspicture.jpeg
|Country=Sweden
|Office=E509
|Affiliation=Halmstad University/CERES
|Revision Date=2016-10-17
|Subject=Information Technology
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignProjects
|project=PPQ
}}
{{AssignSubjectAreas
|SubjectArea=Learning Systems
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignSubjectAreas
|SubjectArea=Smart Environments
}}
{{AssignSubjectAreas
|SubjectArea=Health Technology
}}
{{AssignExaminer
|Examiner=TV4002 Perspektiv på hälsoinnovation
}}
{{AssignExaminer
|Examiner=TV6001 Utvecklingsprojekt med fokus på hälsoinnovation
}}
{{AssignTeacher
|Teacher=Design of Embedded and Intelligent Systems (15 credits)
}}
{{AssignTeacher
|Teacher=Sensor System (7.5 credits)
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}

Jens Lundström

2016-10-17T07:31:26Z

Jens:

Jens Lundström

2016-10-17T07:29:54Z

Jens:

{{Person
|Family Name=Lundström
|Given Name=Jens
|Title=M.Sc
|Phone=+46 35 167865
|Position=Assistant Professor
|Email=jens.lundstrom@hh.se
|Image=jenspicture.jpeg
|Office=E509
|Affiliation=Halmstad University/CERES
|Revision Date=2016-10-17
|Subject=Information Technology
}}
{{AssignProjects
|project=SA3L - Situation Awareness for Ambient Assisted Living
}}
{{AssignProjects
|project=Centre for Health Technology Halland - HCH
}}
{{AssignProjects
|project=PPQ
}}
{{AssignSubjectAreas
|SubjectArea=Machine Learning
}}
{{AssignSubjectAreas
|SubjectArea=Smart Environments
}}
{{AssignSubjectAreas
|SubjectArea=Health Technology
}}
[[Category:Staff]]


{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}

Exploring, modelling and optimization of home care regions

2016-10-17T06:25:13Z

Jens:

{{StudentProjectTemplate
|Summary=This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.
|Keywords=Computational Intelligence, Spatio-Temporal modelling, Clustering, Regression Analysis, Decision Trees, Heteregoenous Data Analysis.
|TimeFrame=Winter 2016 / Spring 2017
|References=Madigan, E. A., & Curet, O. L. (2006). A data mining approach in home healthcare: outcomes and service use. BMC health services research, 6(1), 1.

Cheng, B. W., Chang, C. L., & Liu, I. S. (2005). Enhancing care services quality of nursing homes using data mining. Total Quality Management & Business Excellence, 16(5), 575-596.

Hirdes, J. P., Poss, J. W., & Curtin-Telegdi, N. (2008). The Method for Assigning Priority Levels (MAPLe): a new decision-support system for allocating home care resources. BMC medicine, 6(1), 1.

Harrington, C., Zimmerman, D., Karon, S. L., Robinson, J., & Beutel, P. (2000). Nursing home staffing and its relationship to deficiencies. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 55(5), S278-S287.
|Prerequisites=Courses preferable: Learning Systems, Data Mining. Preferable programming skills: R, Python, Matlab.
|Supervisor=Jens Lundström, Wagner O. De Morais
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
The world is facing a rapid increase of healthcare costs, parts of the reason is due to the growth of the elderly population. Reports from the UN project that the number of people older than 60 years will near quadruple until the year 2050. In Sweden, the number of elderly receiving home-based healthcare service is as well steadily increasing, one approach to partly meet the increasing demand of home-based healthcare is to understand and assist the design of such healthcare by ICT. This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.

The home healthcare management in the municipality of Halmstad, Sweden is divided into approximately 20 regions where each region has different resources, customer satisfaction, costs and residents/customer with different levels of burden of care. Currently, it is desired to support the decisions on how these regions are designed using a data-driven approach. Therefore, this project is about how to explore, understand and model health care region characteristics – for the end goal of providing a decision support system.

Data provided in the project comes from heterogeneous sources such as visiting times, decisions, interventions and positions not necessarily in the same resolution, numerical format or context. This is a challenge - the data can not be directly used by traditional machine learning methods. Moreover is it important to focus on models that are transparent and able to provide insight into how health care regions are modelled.
Research questions include: How to combine the different heterogeneous sources of information? Which machine learning principles/models should be used for learning models representing health care region characteristics? How to interpret such models?

7 workpackages are suggested:

* Literature review as well as exploring how the current (not entirely data-driven) model for designing a home care area is adopted.
* Writing a project plan.
* Writing a data analysis plan and (possible) data collection.
* Explorative data analysis.
* Modelling health care regions (including comparisons of different methods).
* Exploring/understanding/optimizing models for health care regions.
* Reporting (thesis, source code, etc.)

The result is expected to include a detailed description of how the current approach is done today in the home care service as well as how the proposed approach could be used for decision support. Moreover is the results expected to include an evaluation of different machine learning methods suitable for the modelling task.

Exploring, modelling and optimization of home care regions

2016-10-17T06:24:05Z

Jens:

{{StudentProjectTemplate
|Summary=This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.
|Keywords=Computational Intelligence, Spatio-Temporal modelling, Clustering, Regression Analysis, Decision Trees, Heteregoenous Data Analysis.
|TimeFrame=Winter 2016 / Spring 2017
|References=Madigan, E. A., & Curet, O. L. (2006). A data mining approach in home healthcare: outcomes and service use. BMC health services research, 6(1), 1.

Cheng, B. W., Chang, C. L., & Liu, I. S. (2005). Enhancing care services quality of nursing homes using data mining. Total Quality Management & Business Excellence, 16(5), 575-596.

Hirdes, J. P., Poss, J. W., & Curtin-Telegdi, N. (2008). The Method for Assigning Priority Levels (MAPLe): a new decision-support system for allocating home care resources. BMC medicine, 6(1), 1.

Harrington, C., Zimmerman, D., Karon, S. L., Robinson, J., & Beutel, P. (2000). Nursing home staffing and its relationship to deficiencies. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 55(5), S278-S287.
|Prerequisites=Courses preferable: Learning Systems, Data Mining. Preferable programming skills: R, Python, Matlab.
|Supervisor=Jens Lundström, Wagner O. De Morais
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
The world is facing a rapid increase of healthcare costs, parts of the reason is due to the growth of the elderly population. Reports from the UN project that the number of people older than 60 years will near quadruple until the year 2050. In Sweden, the number of elderly receiving home-based healthcare service is as well steadily increasing, one approach to partly meet the increasing demand of home-based healthcare is to understand and assist the design of such healthcare by ICT. This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.
The home healthcare management in the municipality of Halmstad, Sweden is divided into approximately 20 regions where each region has different resources, customer satisfaction, costs and residents/customer with different levels of burden of care. Currently, it is desired to support the decisions on how these regions are designed using a data-driven approach. Therefore, this project is about how to explore, understand and model health care region characteristics – for the end goal of providing a decision support system.

Data provided in the project comes from heterogeneous sources such as visiting times, decisions, interventions and positions not necessarily in the same resolution, numerical format or context. This is a challenge - the data can not be directly used by traditional machine learning methods. Moreover is it important to focus on models that are transparent and able to provide insight into how health care regions are modelled.
Research questions include: How to combine the different heterogeneous sources of information? Which machine learning principles/models should be used for learning models representing health care region characteristics? How to interpret such models?

7 workpackages are suggested:

* Literature review as well as exploring how the current (not entirely data-driven) model for designing a home care area is adopted.
* Writing a project plan.
* Writing a data analysis plan and (possible) data collection.
* Explorative data analysis.
* Modelling health care regions (including comparisons of different methods).
* Exploring/understanding/optimizing models for health care regions.
* Reporting (thesis, source code, etc.)

The result is expected to include a detailed description of how the current approach is done today in the home care service as well as how the proposed approach could be used for decision support. Moreover is the results expected to include an evaluation of different machine learning methods suitable for the modelling task.

Exploring, modelling and optimization of home care regions

2016-10-17T06:23:34Z

Jens: Created page with "{{StudentProjectTemplate |Summary=This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology..."

{{StudentProjectTemplate
|Summary=This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.
|Keywords=Computational Intelligence, Spatio-Temporal modelling, Clustering, Regression Analysis, Decision Trees, Heteregoenous Data Analysis.
|TimeFrame=Winter 2016 / Spring 2017
|References=Madigan, E. A., & Curet, O. L. (2006). A data mining approach in home healthcare: outcomes and service use. BMC health services research, 6(1), 1.

Cheng, B. W., Chang, C. L., & Liu, I. S. (2005). Enhancing care services quality of nursing homes using data mining. Total Quality Management & Business Excellence, 16(5), 575-596.

Hirdes, J. P., Poss, J. W., & Curtin-Telegdi, N. (2008). The Method for Assigning Priority Levels (MAPLe): a new decision-support system for allocating home care resources. BMC medicine, 6(1), 1.

Harrington, C., Zimmerman, D., Karon, S. L., Robinson, J., & Beutel, P. (2000). Nursing home staffing and its relationship to deficiencies. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 55(5), S278-S287.
|Prerequisites=Courses preferable: Learning Systems, Data Mining. Preferable programming skills: R, Python, Matlab.
|Supervisor=Jens Lundström (Primary), Wagner O. De Morais
|Examiner=Antanas Verikas
|Level=Master
|Status=Open
}}
The world is facing a rapid increase of healthcare costs, parts of the reason is due to the growth of the elderly population. Reports from the UN project that the number of people older than 60 years will near quadruple until the year 2050. In Sweden, the number of elderly receiving home-based healthcare service is as well steadily increasing, one approach to partly meet the increasing demand of home-based healthcare is to understand and assist the design of such healthcare by ICT. This project is about developing tools and methods for optimization of health care resources using machine learning as the central technology.
The home healthcare management in the municipality of Halmstad, Sweden is divided into approximately 20 regions where each region has different resources, customer satisfaction, costs and residents/customer with different levels of burden of care. Currently, it is desired to support the decisions on how these regions are designed using a data-driven approach. Therefore, this project is about how to explore, understand and model health care region characteristics – for the end goal of providing a decision support system.

Data provided in the project comes from heterogeneous sources such as visiting times, decisions, interventions and positions not necessarily in the same resolution, numerical format or context. This is a challenge - the data can not be directly used by traditional machine learning methods. Moreover is it important to focus on models that are transparent and able to provide insight into how health care regions are modelled.
Research questions include: How to combine the different heterogeneous sources of information? Which machine learning principles/models should be used for learning models representing health care region characteristics? How to interpret such models?

7 workpackages are suggested:

* Literature review as well as exploring how the current (not entirely data-driven) model for designing a home care area is adopted.
* Writing a project plan.
* Writing a data analysis plan and (possible) data collection.
* Explorative data analysis.
* Modelling health care regions (including comparisons of different methods).
* Exploring/understanding/optimizing models for health care regions.
* Reporting (thesis, source code, etc.)

The result is expected to include a detailed description of how the current approach is done today in the home care service as well as how the proposed approach could be used for decision support. Moreover is the results expected to include an evaluation of different machine learning methods suitable for the modelling task.

Smart Home Simulation

2016-10-16T10:20:51Z

Jens:

{{StudentProjectTemplate
|Summary=Developing and evaluation of a smart home simulator and outlier detection methods.
|Keywords=Ambient assisted living;Intelligent homes;Situation awareness;Machine learning;Outlier detection algorithms
|TimeFrame=Spring 2015
|References=Teresa Garcia-Valverde, Francisco Campuzano, Emilio Serrano, Ana Villa, and Juan A. Botia. 2012. Simulation of human behaviours for the validation of Ambient Intelligence services: A methodological approach. J. Ambient Intell. Smart Environ. 4, 3 (August 2012), 163-181.

Juan A. Botia, Ana Villa, Jose Palma, Ambient Assisted Living system for in-home monitoring of healthy independent elders, Expert Systems with Applications, Volume 39, Issue 9, July 2012, Pages 8136-8148.

Pavel, M.; Jimison, H.B.; Wactlar, H.D.; Hayes, T.L.; Barkis, W.; Skapik, J.; Kaye, J., "The Role of Technology and Engineering Models in Transforming Healthcare," Biomedical Engineering, IEEE Reviews in , vol.6, no., pp.156,177, 2013.
|Prerequisites=Learning Systems
|Supervisor=Jens Lundström, Antanas Verikas, Sławomir Nowaczyk,
|Author=Solved by internal/external resources
|Level=Master
|Status=Finished
|Title=Smart Home Simulation
}}
'''Background'''

Currently, in the area of intelligent homes research, awareness of human behavior is an important research area. This area has for the last decades included non-trivial problems such as: activity recognition (what the person is doing), fall monitoring and in-door tracking of persons. However, the data acquisition phase for this kind of research is often a time-consuming, protracted and an expensive process which often is followed by the workload of maintaining the data set such as handling missing values. A complementary tool that helps in the process is a smart home simulator that is able to output sensor sequences similar to real sensor sequences. Such tool could be used for generating new data sets from models computed from previously collected data. Moreover, such generated sequences could be injected with anomalous sensor sequences to study the effect in several aspects such as:

* Algorithms for outlier detection.
* Algorithms for classifying different types of users.
* Algorithms for visualizing data and analysis results.

'''Project description'''

The overall goal in this project is to develop a GUI-based intelligent home simulator able to output realistic sensor sequences given prior knowledge such as the planning of the home environment (possible from a CAD-file format), sensor positions and normal behavior of a person. The secondary, yet important goal, is to study, implement and validate algorithms for outlier detection using the generated data. The outlier detection algorithm should be compatible with the simulator previously developed at IS-lab.

'''Activity plan'''

The suggested project could be specified in the following work packages:

'''WP1'''
*Studying related work within the field of simulation, smart homes and outlier detection.
*Writing a project plan.

'''WP2'''
*Develop a GUI-based smart home simulator in an appropriate programming language (or using an existing simulation tool).
*Study and generate realistic anomalous sequences

'''WP3'''
*Studying, implementing and evaluating outlier detection methods able to detect injected anomalous sensor sequences.

Investigating Robustness of DNNs

2016-10-16T10:16:11Z

Jens:

{{StudentProjectTemplate
|Summary=This master thesis project aims at characterizing sensitivity to classification of images (based on deep neural networks).
|Keywords=deep neural networks, robustness
|TimeFrame=Spring 2015
|References=Szegedy, Christian, et al. "Intriguing properties of neural networks." arXiv preprint arXiv:1312.6199 (2013).

Hinton, Geoffrey E., and Ruslan R. Salakhutdinov. "Reducing the dimensionality of data with neural networks." Science 313.5786 (2006): 504-507.

Hinton, Geoffrey E. "Learning multiple layers of representation." Trends in cognitive sciences 11.10 (2007): 428-434.

Nguyen, Anh, Jason Yosinski, and Jeff Clune. "Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images." arXiv preprint arXiv:1412.1897 (2014).

Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.
|Prerequisites=Learning Systems, Data Mining, Parallel programming
|Supervisor=Jens Lundström, Stefan Byttner,
|Examiner=Antanas Verikas
|Author=Matej Uličný
|Level=Master
|Status=Finished
}}
Deep Neural Networks (DNNs) have gained much interest during the last years. Among many sucessful applications, DNNs have shown outstanding performance in the task of learning feature representations and classification of images. A state-of-art, high accurate, neural network trained to classify 1.2 million images using 60 million parameters and 650,000 neurons was developed by Hinton et al. (2012). However, recent findings reveal delicate difficulties on noise robustness in DNNs, Szegedy et al (2013). The purpose of the thesis is to characterize the sensitivity of DNNs and potentially make suggestions on how robustness can be achieved.
The thesis project aims at two related studies. Firstly, the master student will investigate how meaningsless, to human, images are classified with high confidence using DNNs, as reported by other studies. Secondly, the student will investige DNNs misclassifications of images with small pertubations, not visible to humans. Moreover, the student is also encouraged to apply image preprocessing methods in order to increase classification accuracy.

Four work packages are suggested:

1. Background study on DNNs and related reserarch.
2. Practical tests on DNNs on medium size datasets.
3. Investigation of distorted (meaningless) images classified with high confidence.
4. Investigation of misclassifications of images with small pertubations not visible to humans.

The result is expected to include investigation results and conclusions on both of the concerned research questions described above.

Investigating Robustness of DNNs

2014-12-17T08:53:33Z

Jens:

{{StudentProjectTemplate
|Summary=This master thesis project aims at characterizing sensitivity to classification of images (based on deep neural networks).
|Keywords=deep neural networks, robustness
|TimeFrame=Spring 2015
|References=Szegedy, Christian, et al. "Intriguing properties of neural networks." arXiv preprint arXiv:1312.6199 (2013).

Hinton, Geoffrey E., and Ruslan R. Salakhutdinov. "Reducing the dimensionality of data with neural networks." Science 313.5786 (2006): 504-507.

Hinton, Geoffrey E. "Learning multiple layers of representation." Trends in cognitive sciences 11.10 (2007): 428-434.

Nguyen, Anh, Jason Yosinski, and Jeff Clune. "Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images." arXiv preprint arXiv:1412.1897 (2014).

Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.
|Prerequisites=Learning Systems, Data Mining, Parallel programming
|Supervisor=Jens Lundström, Stefan Byttner,
|Level=Master
|Status=Open
}}
Deep Neural Networks (DNNs) have gained much interest during the last years. Among many sucessful applications, DNNs have shown outstanding performance in the task of learning feature representations and classification of images. A state-of-art, high accurate, neural network trained to classify 1.2 million images using 60 million parameters and 650,000 neurons was developed by Hinton et al. (2012). However, recent findings reveal delicate difficulties on noise robustness in DNNs, Szegedy et al (2013). The purpose of the thesis is to characterize the sensitivity of DNNs and potentially make suggestions on how robustness can be achieved.
The thesis project aims at two related studies. Firstly, the master student will investigate how meaningsless, to human, images are classified with high confidence using DNNs, as reported by other studies. Secondly, the student will investige DNNs misclassifications of images with small pertubations, not visible to humans. Moreover, the student is also encouraged to apply image preprocessing methods in order to increase classification accuracy.

Four work packages are suggested:

1. Background study on DNNs and related reserarch.
2. Practical tests on DNNs on medium size datasets.
3. Investigation of distorted (meaningless) images classified with high confidence.
4. Investigation of misclassifications of images with small pertubations not visible to humans.

The result is expected to include investigation results and conclusions on both of the concerned research questions described above.

Investigating Robustness of DNNs

2014-12-17T07:34:50Z

Jens:

{{StudentProjectTemplate
|Summary=This master thesis project aims at characterizing sensitivity to classification of images (based on deep neural networks).
|Keywords=deep neural networks, robustness
|TimeFrame=Spring 2015
|References=Szegedy, Christian, et al. "Intriguing properties of neural networks." arXiv preprint arXiv:1312.6199 (2013).

Hinton, Geoffrey E., and Ruslan R. Salakhutdinov. "Reducing the dimensionality of data with neural networks." Science 313.5786 (2006): 504-507.

Hinton, Geoffrey E. "Learning multiple layers of representation." Trends in cognitive sciences 11.10 (2007): 428-434.

Nguyen, Anh, Jason Yosinski, and Jeff Clune. "Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images." arXiv preprint arXiv:1412.1897 (2014).

Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.
|Prerequisites=Learning Systems, Data Mining, Parallel programming
|Supervisor=Jens Lundström, Stefan Byttner,
|Level=Master
|Status=Open
}}
Deep Neural Networks (DNNs) have gained much interest during the last years. Among many sucessful applications, DNNs have shown outstanding performance in the task of learning feature representations and classification of images. A state-of-art, high accurate, neural network trained to classify 1.2 million images using 60 million parameters and 650,000 neurons was developed by Hinton et al. (2012). However, recent findings reveal delicate difficulties on noise robustness in DNNs, Szegedy et al (2013). This master thesis project aims at two related studies. Firstly, the master student will investigate how meaningsless, to human, images are classified with high confidence using DNNs, as reported by other studies. Secondly, the student will investige DNNs misclassifications of images with small pertubations, not visible to humans. Moreover, the student is also encouraged to apply image preprocessing methods in order to increase classification accuracy.

Four work packages are suggested:

1. Background study on DNNs and related reserarch.
2. Practical tests on DNNs on medium size datasets.
3. Investigation of distorted (meaningless) images classified with high confidence.
4. Investigation of misclassifications of images with small pertubations not visible to humans.

The result is expected to include investigation results and conclusions on both of the concerned research questions described above.

Investigating Robustness of DNNs

2014-12-17T06:25:01Z

Jens: Created page with "{{StudentProjectTemplate |Summary=This master thesis project aims at characterizing sensitivity to classification of images (based on deep neural networks). |Keywords=deep neu..."

{{StudentProjectTemplate
|Summary=This master thesis project aims at characterizing sensitivity to classification of images (based on deep neural networks).
|Keywords=deep neural networks, robustness
|TimeFrame=Spring 2015
|Prerequisites=Learning Systems, Data Mining
|Supervisor=Jens Lundström, Stefan Byttner,
|Level=Master
|Status=Open
}}