ISLAB/CAISR - User contributions [en]

Data analysis in collaboration with WirelessCar

2020-10-05T13:45:16Z

Peyman:

{{StudentProjectTemplate
|Summary=Data analysis in collaboration with WirelessCar
|Supervisor=Mahmoud Rahat, Peyman Mashhadi, Slawomir Nowaczyk
|Level=Master
|Status=Open
}}
WirelessCar is a very prospective company with the main mantra of leading the automotive industry towards a digital society. We are provided with an interesting dataset from the WirelessCar company. The goal of this thesis is twofold. The first goal targets advanced exploratory data analysis (EDA) to gain insight from data and ideally convert it into a story. After gaining insight from the data, the next step is to define an exciting application with both research and business values. From the machine learning point of view, this application could be anything from supervised learning, unsupervised learning, feature representation learning, adversarial learning, and so on. The real advantage of this project is that you get to work with a real dataset and going through the entire pipeline of designing a successful project from both research and business perspective, in close collaboration with WirelessCar.

The dataset contains information about many trips taken by different vehicles. It is represented in a hierarchy with three levels. The first level is the coarse-grained information about each trip, including start and end GPS position, total fuel consumption, and more high-level information. The second level includes a more detailed representation of each trip. The third level breaks down each trip into different segments called waypoints and contains information sampled for each segment. There are many use cases that WirelessCar could be interested in. To name a few, they are interested in finding some driver-related behavior patterns and their relation to cost, time, and traffic. One interesting question will be if the route taken by a driver to a specific location is cost-optimal? This can be addressed using EDA by comparing it to other drivers’ behaviors. This can be taken to the next step and be accompanied by a machine learning algorithm predicting optimal route taking into account time and cost. Another use case would be the possibility of carpooling and consequently reducing traffic and cost. These use cases can be well-understood and analyzed by advanced exploratory data analysis and further be converted into a high-impact machine learning problem.

Data analysis in collaboration with WirelessCar

2020-10-05T13:44:12Z

Peyman:

{{StudentProjectTemplate
|Summary=Details to be decided...
|Supervisor=Mahmoud Rahat, Peyman Mashhadi, Slawomir Nowaczyk
|Level=Master
|Status=Draft
}}
WirelessCar is a very prospective company with the main mantra of leading the automotive industry towards a digital society. We are provided with an interesting dataset from the WirelessCar company. The goal of this thesis is twofold. The first goal targets advanced exploratory data analysis (EDA) to gain insight from data and ideally convert it into a story. After gaining insight from the data, the next step is to define an exciting application with both research and business values. From the machine learning point of view, this application could be anything from supervised learning, unsupervised learning, feature representation learning, adversarial learning, and so on. The real advantage of this project is that you get to work with a real dataset and going through the entire pipeline of designing a successful project from both research and business perspective, in close collaboration with WirelessCar.

The dataset contains information about many trips taken by different vehicles. It is represented in a hierarchy with three levels. The first level is the coarse-grained information about each trip, including start and end GPS position, total fuel consumption, and more high-level information. The second level includes a more detailed representation of each trip. The third level breaks down each trip into different segments called waypoints and contains information sampled for each segment. There are many use cases that WirelessCar could be interested in. To name a few, they are interested in finding some driver-related behavior patterns and their relation to cost, time, and traffic. One interesting question will be if the route taken by a driver to a specific location is cost-optimal? This can be addressed using EDA by comparing it to other drivers’ behaviors. This can be taken to the next step and be accompanied by a machine learning algorithm predicting optimal route taking into account time and cost. Another use case would be the possibility of carpooling and consequently reducing traffic and cost. These use cases can be well-understood and analyzed by advanced exploratory data analysis and further be converted into a high-impact machine learning problem.

Reinforcement Learning with Adaptive Representation Learning

2020-10-05T13:38:39Z

Peyman: This project targets finding representations that make the reinforcement learning more efficient in terms of finding an easier state to action mapping.

{{StudentProjectTemplate
|Summary=This project targets finding representations that make the reinforcement learning more efficient in terms of finding an easier state to action mapping.
|Keywords=Reinforcement Learning, Representation Learning, Deep Learning
|References=IS A GOOD REPRESENTATION SUFFICIENT FOR SAMPLE EFFICIENT REINFORCEMENT LEARNING?, Simon S. Du, Sham M. Kakade, 2020

Learning State Representations for Query Optimization with Deep Reinforcement Learning, Jennifer Ortiz, Magdalena Balazinska, Johannes Gehrke, S. Sathiya Keerthi, 2018

State Representation Learning for Control: An Overview, Timothée Lesort, Natalia Díaz-Rodríguez, Jean-François Goudou, and David Filliat, 2018
|Supervisor=Alexander Galozy , Peyman Mashhadi
|Level=Master
|Status=Open
}}
This project targets finding representations that make the reinforcement learning more efficient in terms of finding an easier state to action mapping. As a concrete example, let’s take the task of fitness of a person, and assume that the data is received in the form of images. Images are high dimensional data which can take many different states. This large state space would make it difficult to find an optimal action for the task in a reasonable amount of time. However, let’s imagine that we could convert those images into another representation that extract certain features like weight, highs, muscle mass, and similar important features for fitness evaluation. If we could find those features, then finding optimal actions would be much easier. The goal of this project is actually to learn the representation of incoming data in a sequential manner into a much simpler and more informative representation for the task at hand.

In reinforcement learning, most of the time, the state representation and actions are fixed and only the probabilities of the right action given the current state are changed over time. However, in this research, the representation is subject to being updated, as we learn what features are more important for solving a task. As a concrete example, one way to approach it is to have a have an attention mechanism on the features, selectively taking features into account that maximize cumulative reward. Another approach could be to have an encoder and transform the representations into bottleneck representation which can provide the new states. Then, based on the action in the new state-action space, a new reward will be calculated and the reward is used to backpropagate and update the representation.

Reinforcement Learning with Adaptive Representation Learning

2020-10-05T13:35:13Z

Peyman: Created page with "{{StudentProjectTemplate |Summary=In reinforcement learning, most of the time, the state representation and actions are fixed and only the probabilities of the right action gi..."

{{StudentProjectTemplate
|Summary=In reinforcement learning, most of the time, the state representation and actions are fixed and only the probabilities of the right action given the current state is changed over time. However, in this research, the representation is subject to being updated, as we learn what features are more important for solving a task.
|Keywords=Reinforcement Learning, Representation Learning, Deep Learning
|References=IS A GOOD REPRESENTATION SUFFICIENT FOR SAMPLE EFFICIENT REINFORCEMENT LEARNING?, Simon S. Du, Sham M. Kakade, 2020

Learning State Representations for Query Optimization with Deep Reinforcement Learning, Jennifer Ortiz, Magdalena Balazinska, Johannes Gehrke, S. Sathiya Keerthi, 2018

State Representation Learning for Control: An Overview, Timothée Lesort, Natalia Díaz-Rodríguez, Jean-François Goudou, and David Filliat, 2018
|Supervisor=Alexander Galozy , Peyman Mashhadi
|Level=Master
|Status=Open
}}
This project targets finding representations that make the reinforcement learning more efficient in terms of finding an easier state to action mapping. As a concrete example, let’s take the task of fitness of a person, and assume that the data is received in the form of images. Images are high dimensional data which can take many different states. This large state space would make it difficult to find an optimal action for the task in a reasonable amount of time. However, let’s imagine that we could convert those images into another representation that extract certain features like weight, highs, muscle mass, and similar important features for fitness evaluation. If we could find those features, then finding optimal actions would be much easier. The goal of this project is actually to learn the representation of incoming data in a sequential manner into a much simpler and more informative representation for the task at hand.

In reinforcement learning, most of the time, the state representation and actions are fixed and only the probabilities of the right action given the current state are changed over time. However, in this research, the representation is subject to being updated, as we learn what features are more important for solving a task. As a concrete example, one way to approach it is to have a have an attention mechanism on the features, selectively taking features into account that maximize cumulative reward. Another approach could be to have an encoder and transform the representations into bottleneck representation which can provide the new states. Then, based on the action in the new state-action space, a new reward will be calculated and the reward is used to backpropagate and update the representation.

Multitask learning on vehicle data

2020-09-28T14:14:02Z

Peyman: Created page with "{{StudentProjectTemplate |Summary=Learning shared representation using multitask learning on a vehicle-related data |Keywords=Multitask learning, Transfer learning, Shared rep..."

{{StudentProjectTemplate
|Summary=Learning shared representation using multitask learning on a vehicle-related data
|Keywords=Multitask learning, Transfer learning, Shared representation, Vehicle data
|Supervisor=Mahmoud Rahat, Peyman Mashhadi
|Level=Master
|Status=Open
}}
Multitask learning is one approach to address transfer learning. It uses information contained in the training signal of related tasks. Multitask learning improves performance and generalization by finding a part of feature space or transformed feature space useful for all the related tasks. To achieve this shared representation, all the related tasks are trained in parallel. From another perspective, multitask learning can be viewed as a regularization technique due to the imposed requirement of shared representation appropriate for all the related tasks. This form of regularization can be superior to other regularizers that penalize overfitting or complexity of the models.

The goal of this Master’s thesis proposal is to adopt multitask learning on a vehicle-related dataset. That could include many applications such as fuel consumption, predictive maintenance, and etc.

MMultitask Learning on Vehicle Data

2020-09-28T14:08:20Z

Peyman: Created page with "{{StudentProjectTemplate |Summary=Learning shared representation using multitask learning on a vehicle-related data |Supervisor=Mahmoud Rahat, Peyman Mashhadi }} Multitask lea..."

{{StudentProjectTemplate
|Summary=Learning shared representation using multitask learning on a vehicle-related data
|Supervisor=Mahmoud Rahat, Peyman Mashhadi
}}
Multitask learning is one approach to address transfer learning. It uses information contained in the training signal of related tasks. Multitask learning improves performance and generalization by finding a part of feature space or transformed feature space useful for all the related tasks. To achieve this shared representation, all the related tasks are trained in parallel. From another perspective, multitask learning can be viewed as a regularization technique due to the imposed requirement of shared representation appropriate for all the related tasks. This form of regularization can be superior to other regularizers that penalize overfitting or complexity of the models.

The goal of this Master’s thesis proposal is to adopt multitask learning on a vehicle-related dataset. That could include many applications such as fuel consumption, predictive maintenance, and etc.

Deep stacked ensemble

2019-10-02T09:31:47Z

Peyman:

{{StudentProjectTemplate
|Summary=This project aims at training multiple parallel deep networks in such a way to learn different representation of data which will be suitable to frame these networks in stacked ensemble framework.
|Keywords=Deep learning, Staked ensemble, Statistics
|TimeFrame=Fall 2019
|References=1- David H.Wolpert, "Stacked generalisation" https://doi.org/10.1016/S0893-6080(05)80023-1

2- Jason Brownle, "How to Develop a Stacking Ensemble for Deep Learning Neural Networks in Python With Keras", https://machinelearningmastery.com/stacking-ensemble-for-deep-learning-neural-networks/
|Prerequisites=deep learning, data mining,
programming knowledge of one of deep learning frameworks such as tensorflow, pytorch or at least their APIs
|Supervisor=Sławomir Nowaczyk, Peyman Mashhadi
|Level=Master
|Status=Open
}}
Stacking is a form of ensemble learning that combines multiple models through a meta model. In its basic form it is made up of two layers: base layer and meta layer. The base layer models is trained on the original features of dataset, while the meta model consumes predictions of the base models to generate the final predictions. Stacking has won many prestigious machine learning competitions.

One important fact is that the meta model performs well when the base models have acceptable performances and at the same time have low correlations to each other. Currently, there is no automatic approach for selection the base models' structures. It is basically done based on trial and error and based on prior experiences and knowledge.

The aim of this project is to provide an integrated automatic stacking model in a deep learning fashion. This integrated stacked deep net is comprised of multiple parallel deep nets (with the exact same structure) which are followed by another network. The first part (parallel part) plays the role of base model, and the rest of the structure after parallel part plays the role of meta model. The idea is to train this structure in a way that each parallel network learn different representations of the data at the level of parallel part so that the meta model can take advantage of their low correlated predictions.

Deep stacked ensemble

2019-10-02T09:02:01Z

Peyman: Created page with "{{StudentProjectTemplate |Summary=This project aims at training multiple parallel deep networks in such a way to learn different representation of data which will be suitable ..."

{{StudentProjectTemplate
|Summary=This project aims at training multiple parallel deep networks in such a way to learn different representation of data which will be suitable to frame these networks in stacked ensemble framework.
|Keywords=Deep learning, Staked ensemble, Statistics
|TimeFrame=Fall 2019
|References=1- David H.Wolpert, "Stacked generalisation" https://doi.org/10.1016/S0893-6080(05)80023-1

2- Jason Brownle, "How to Develop a Stacking Ensemble for Deep Learning Neural Networks in Python With Keras", https://machinelearningmastery.com/stacking-ensemble-for-deep-learning-neural-networks/
|Prerequisites=Deep learning,
|Supervisor=Sławomir Nowaczyk, Peyman Mashhadi
|Level=Master
|Status=Open
}}
Stacking is a form of ensemble learning that combines multiple models through a meta model. In its basic form it is made up of two layers: base layer and meta layer. The base layer models is trained on the original features of dataset, while the meta model consumes predictions of the base models to generate the final predictions. Stacking has won many prestigious machine learning competitions.

One important fact is that the meta model performs well when the base models have acceptable performances and at the same time have low correlations to each other. Currently, there is no automatic approach for selection the base models' structures. It is basically done based on trial and error and based on prior experiences and knowledge.

The aim of this project is to provide an integrated automatic stacking model in a deep learning fashion. This integrated stacked deep net is comprised of multiple parallel deep nets (with the exact same structure) which are followed by another network. The first part (parallel part) plays the role of base model, and the rest of the structure after parallel part plays the role of meta model. The idea is to train this structure in a way that each parallel network learn different representations of the data at the level of parallel part so that the meta model can take advantage of their low correlated predictions.