Several statistical problems can be described as estimation problem, where the goal is to learn a set of parameters, from some data, by maximizing a criterion. These type of problems are typically encountered in a supervised learning setting, where we want to relate an output (or many outputs) to multiple inputs. The relationship between these outputs and these inputs can be complex, and this complexity can be attributed to the high dimensionality of the space containing the inputs and the outputs; the existence of a structural prior knowledge within the inputs or the outputs that if ignored may lead to inefficient estimates of the parameters; and the presence of a non-trivial noise structure in the data. In this thesis we propose new statistical methods to achieve model selection and estimation when there are more predictors than observations. We also design a new set of algorithms to efficiently solve the proposed statistical models. We apply the implemented methods to genetic data sets of cancer patients and to some economics data.

Specifically equipped and programmed robots are highly successful in controlled industrial environments such as automated production lines. For the transition of robots from such controlled uniform environments to unconstrained household environments with a large range of conditions and variations, a new paradigm is needed to prepare the robots for deployment. Robots need to be able to quickly adapt to their changing environments and learn on their own how to solve their tasks in novel situations. This dissertation focusses on the aspect of learning to predict the success of two-object means-end actions in a developmental way. E.g. the action of bringing one object into reach by pulling another, where the one object is on top of the other. Here it is the “on top” relation that affects the success of the action. Learning the preconditions for complex means-end actions via supervised learning can take several thousand training samples, which is impractical to generate, hence more rapid learning capabilities are necessary. Three contributions of this dissertation are used to solve the learning problem. 1. Inspired by infant psychology this dissertation investigates an approach to intrinsic motivation that is based on active learning, guiding the robot's exploration to create experience useful for improving classification performance. 2. This dissertation introduces histogram based 3D vision features that encode the relative spatial relations between surface points of object pairs, allowing a robot to reliably recognise the important spatial categories that affect means-end action outcomes. 3. Intrinsically encoded experience is extracted into symbolic category knowledge, encoding higher level abstract categories. These symbolic categories are used for knowledge transfer by using them to extend the state space of action precondition learning classifiers. Depending on the actions and their preconditions, the contributions of this dissertation enable a robot to achieve success prediction accuracies above 85% with ten training samples instead of approximately 1000 training samples that would otherwise be required. These results can be achieved when (a) the action preconditions can be easily identified from the used vision features or (b) the action preconditions to be learnt rest upon already existing knowledge, then it is possible to achieve these results by reusing the existing knowledge. This dissertation demonstrates, in simulation, an alternative to handcoding the knowledge required for a robot to interact with and manipulate objects in the environment. It shows that rapid learning, grounded in autonomous exploration, can be feasible if the necessary vision features are constructed and if existing knowledge is consistently reused.

Thesis made openly available per email from author, 8-2-2015.

published_or_final_version / Mechanical Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / Computer Science / Doctoral / Doctor of Philosophy

published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Mathematics / Doctoral / Doctor of Philosophy

Data on the three?dimensional (3?D) morphology of normal faces and facial deformities as well as objective techniques for evaluating postoperative changes are currently unavailable. With the advent of stereophotography and cone?beam CT (CBCT), it is possible to acquire 3?D images of soft and hard tissues of the maxillofacial complex. A series of studies were conducted aimed at 1) establishing 3?D facial anthropometric norms for Hong Kong young adults, 2) determining 3?D facial anthropometric features in skeletal Class II and III deformities, 3) characterizing the oropharyngeal space in Class II and III skeletal deformities, 4) exploring clinical applications of 3?D colour maps, 5) establishing a non?invasive technique for estimating serial volume changes and 6) creating virtual craniofacial models by fusing 3?D photographs and CBCT images. Study 1: A stereophotographic system was used to capture 3?D images of 103 Hong Kong Chinese young adults with normal balanced faces. An anthropometric analysis protocol with linear, angular and proportional measurements was developed to establish a normative database and quantify dysmorphology. The Hong Kong Chinese norms were distinct from Caucasians, especially with regard to ocular and nasal measurements. Facial height and nasolabial measurements differed significantly between Hong Kong males and females. Study 2: Anthropometric analyses of 3?D facial images from 41 skeletal Class II and 43 Class III subjects were performed. The Class II subjects had increased lower facial height compared with Class III, who had longer total facial heights and narrower faces. While Class II deformity primarily resulted from mandibular deficiency with a normal maxilla, Class III presented as combined midfacial hypoplasia and mandibular hyperplasia. Study 3: Anthropometric characteristics of the oropharygeal space in skeletal Class II and III were evaluated using 62 CBCT scans. The retroglossal (RG) and retropalatal (RP) volumes and average cross sectional areas were significantly larger in Class III than Class II skeletal deformity. The RP compartment was larger but less uniform than the RG compartment in both Classes. Study 4: 3?D photographs or CBCT images acquired at two different time points were superimposed using a common unaffected area. 3?D colour maps were generated depicting distance differences between superimposed images in a graphical format. These maps were used as an objective tool for treatment planning and assessing outcomes after orthognathic surgery, bimaxillary distraction and facial trauma. Study 5: 3?D photogrammetry was employed for planning soft tissue expansion (STE) and transplantation of a vascularised scapular flap in hemifacial microsomia. This technique facilitated the identification of extent and degree of tissue deficiency, selection of the appropriate tissue expander, monitoring volumetric changes during STE and estimation of the free flap dimensions. Study 6: 3?D facial photographs and CBCT scans of 29 subjects were merged to create virtual craniofacial models with natural surface texture. Accuracy was assessed with 3?D colour maps and Root Mean Square (RMS) error. The CBCT and 3?D photographic data were integrated while minimizing average RMS error to 0.441mm. These virtual composite craniofacial models permitted concurrent 3?D assessment of bone and soft tissue. / published_or_final_version / Dentistry / Doctoral / Doctor of Philosophy

No description available.