Probabilistic study of instantiated gaussian processes and application to spatio-temporal data. September or October Application deadline date: April 14th Decision announcement date: The research will be undertaken in the context of an interdisciplinary project involving also Hubert Curien Laboratory from the University Jean Monnet of St Etienne.
The consortium has scientific expertise on probability and statistics, information and image processing, and machine learning, providing a stimulating scientific environment for this thesis.
Last but not least, St Etienne is a very pleasant place to study and work. St Etienne is rated each year as one of the best place in France for studying. PhD thesis subject Gaussian processes are non-linear models of continuous random processes which are widely used to describe numerical data as sounds, images, videos, etc. A Gaussian process is defined mainly by its expectation function and its covariance function the kernel.
The description of the kernel using parametric functions and the estimation of these parameters form the focus of many recent works [L05,D16]. In the context of image sequences knowing that our study is intended to address other types of data , the main objective is no longer to describe a Gaussian process but a set of Gaussian processes that can possess instances Different temporal or spatial supports , with the aim to analyse videos with dynamic textures lights, waves, clouds, fields of wheat … taken from different angles for example.
The main objective of the thesis is to provide a precise mathematical framework for these instanciated Gaussian processes in order to be able to estimate the different parameters instances, mathematical expectations and kernels' parameters. First, the PhD student will be intended to make a state-of-the-art about the different kernels and their properties, mainly their stationarity in time and space in order to propose new kernels.
The next step is to develop robust parameter estimation methods and to work on the automatic selection of the kernels. Then, the formalism of non-stationary and instanciated Gaussian processes will be developed, together with their numerical simulations. The last step concerns the mixture of instanciated Gaussian processes and their application to real data like videos.
Knowledges in image processing and machine learning would also be appreciated. Application process Your application should include the following documents: On Pattern Analysis and Machine Intelligence, vol.
All the information is available at https: The project and job description We aim to develop the next generation of statistical inference methods to analyze Earth Observation EO data. Machine learning models have helped to monitor land, oceans, and atmosphere through the analysis and estimation of climate and biophysical parameters. Current approaches, however, cannot deal efficiently with the particular characteristics of remote sensing data.
Highly motivated researchers with a degree in computer science, statistics, machine learning, electrical engineering, physics, or mathematics are encouraged to apply. All candidates should have a solid understanding and knowledge of machine learning and statistics, and being particularly interested in remote sensing and geoscience problems.
The thesis will address problems in regression, graphical models and causal inference. Application details - Deadline: Send your application no later than April 1st June - How long?
Salary according to UV scales including social security, health insurance benefits, and travel money - Where? Valencia, Spain, Mediterranean city, nice weather, hike and beach. Informal inquiries may be addressed to Prof. Send your dossier in one single PDF to gustau. Last years have seen the massive adoption of deep learning techniques for various tasks in computer vision. In remote sensing and Earth-observation data analysis, our team has developed algorithms for classification and detection which have established new state-of-the-art performances.
With this new PhD thesis, we now want to discover how deep networks can help understanding the multitemporal satellite image series. Research axis will include: It is therefore not only essential to use our resources as efficiently as possible, but also to assess and mitigate the risk to crops. Farming is increasingly driven by machines and with less staff it is difficult for farmers to monitor their crop.
There is the drive to use airborne technologies as well as satellite imagery to do this remotely and automate this. The project aims to build up an extensible, probabilistic framework to do this resulting in a database and data model. The ultimate aim is to not just have a database of the spectral signatures of different plant species, but also to incorporate phenology and health status of the plants.
With regards to crops this will help mitigate the risks of droughts and diseases. Irrigation can be directed where needed and fertilizer used more effectively. Early intervention can stop diseases spreading and there will be less use of pesticides and fungicides. Being able to choose an optimal time to harvest will lead to less food wastage.
Using sophisticated machine learning techniques, sparse models of the land cover can be created. These models will help where there is limited up and downlink bandwidth as there are with airborne technologies as well as with satellites. The device gathering the data can carry a sparse model and only where new data is significantly different to the model action is necessary.
In the first instance this action will be an alert to an anomaly. Further analysis is then necessary whether the anomaly is expected due to e. The aim of this PhD project is a probabilistic model of spectral signatures of plants incorporating phenology and diseases.
To this end various machine learning techniques will be employed and benchmarked against each other. Data from the CropScape database https: This can then be enriched with data from the Sentinel satellites for temporal analysis, since the revisit times are shorter. Different resolutions and number of spectral bands need to be given consideration. The techniques will also be assessed on their ability to generate knowledge automatically, for example in which way the spectral signature of a plant changes under increasing drought conditions and whether there are underlying general principles.
Another aspect is the development of a confidence measure identifying and quantifying classification error. Applicants should have a masters degree in mathematics or a closely related subject, e. The funds for this studentship are available for 3 years in the first instance. In order to be considered for this studentship please submit a formal application to the PhD in Applied Mathematics and Theoretical Physics, University of Cambridge via the University's Graduate Admissions website for more information on this please visit http: Applications should be submitted online until the 31st of March Expressions of interest letter that briefly describe your motivation for this project should be sent to grad-administrator maths.
Please quote reference LE on your application and in any correspondence about this vacancy. The University values diversity and is committed to equality of opportunity. The Department would particularly welcome applications from women, since women are, and have historically been, underrepresented on our student cohort.
The University has a responsibility to ensure that all employees are eligible to live and work in the UK.