Online self-supervised multi-instance segmentation of dynamic objects

This paper presents a method for the continuous segmentation of dynamic objects using only a vehicle mounted monocular camera without any prior knowledge of the object’s appearance. Prior work in online static/dynamic segmentation is extended to identify multiple instances of dynamic objects by introducing an unsupervised motion clustering step. These clusters are then used to update a multi-class classifier within a self-supervised framework. In contrast to many tracking-by-detection based methods, our system is able to detect dynamic objects without any prior knowledge of their visual appearance shape or location. Furthermore, the classifier is used to propagate labels of the same object in previous frames, which facilitates the continuous tracking of individual objects based on motion. The proposed system is evaluated using recall and false alarm metrics in addition to a new multi-instance labelled dataset to evaluate the performance of segmenting multiple instances of objects.

Self-Supervised ARTMAP

Computational models of learning typically train on labeled input patterns (supervised learning), unlabeled input patterns (unsupervised learning), or a combination of the two (semisupervised learning). In each case input patterns have a fixed number of features throughout training and testing. Human and machine learning contexts present additional opportunities for expanding incomplete knowledge from formal training, via self-directed learning that incorporates features not previously experienced. This article defines a new self-supervised learning paradigm to address these richer learning contexts, introducing a neural network called self-supervised ARTMAP. Self-supervised learning integrates knowledge from a teacher (labeled patterns with some features), knowledge from the environment (unlabeled patterns with more features), and knowledge from internal model activation (self-labeled patterns). Self-supervised ARTMAP learns about novel features from unlabeled patterns without destroying partial knowledge previously acquired from labeled patterns. A category selection function bases system predictions on known features, and distributed network activation scales unlabeled learning to prediction confidence. Slow distributed learning on unlabeled patterns focuses on novel features and confident predictions, defining classification boundaries that were ambiguous in the labeled patterns. Self-supervised ARTMAP improves test accuracy on illustrative lowdimensional problems and on high-dimensional benchmarks. Model code and benchmark data are available from: http://techlab.bu.edu/SSART/.

Contributions to unsupervised and supervised learning with applications in digital image processing

311 p. : il.

Evaluating retraining rules for semi-supervised learning in neural network based cursive word recognition

Training a system to recognize handwritten words is a task that requires a large amount of data with their correct transcription. However, the creation of such a training set, including the generation of the ground truth, is tedious and costly. One way of reducing the high cost of labeled training data acquisition is to exploit unlabeled data, which can be gathered easily. Making use of both labeled and unlabeled data is known as semi-supervised learning. One of the most general versions of semi-supervised learning is self-training, where a recognizer iteratively retrains itself on its own output on new, unlabeled data. In this paper we propose to apply semi-supervised learning, and in particular self-training, to the problem of cursive, handwritten word recognition. The special focus of the paper is on retraining rules that define what data are actually being used in the retraining phase. In a series of experiments it is shown that the performance of a neural network based recognizer can be significantly improved through the use of unlabeled data and self-training if appropriate retraining rules are applied.

Leveraging Web 2.0 data for scalable semi-supervised learning of domain-specific sentiment lexicons

Since manually constructing domain-specific sentiment lexicons is extremely time consuming and it may not even be feasible for domains where linguistic expertise is not available. Research on the automatic construction of domain-specific sentiment lexicons has become a hot topic in recent years. The main contribution of this paper is the illustration of a novel semi-supervised learning method which exploits both term-to-term and document-to-term relations hidden in a corpus for the construction of domain specific sentiment lexicons. More specifically, the proposed two-pass pseudo labeling method combines shallow linguistic parsing and corpusbase statistical learning to make domain-specific sentiment extraction scalable with respect to the sheer volume of opinionated documents archived on the Internet these days. Another novelty of the proposed method is that it can utilize the readily available user-contributed labels of opinionated documents (e.g., the user ratings of product reviews) to bootstrap the performance of sentiment lexicon construction. Our experiments show that the proposed method can generate high quality domain-specific sentiment lexicons as directly assessed by human experts. Moreover, the system generated domain-specific sentiment lexicons can improve polarity prediction tasks at the document level by 2:18% when compared to other well-known baseline methods. Our research opens the door to the development of practical and scalable methods for domain-specific sentiment analysis.

Self-regulated learning and executive function : exploring the relationships in a sample of adolescent males

This study investigated relationships between SRL and EF in a sample of 254 school-aged adolescent males. Two hypotheses were tested: that self-reported measures of SRL and EF are closely related and that as different aspects of EF mature during adolescence, the corresponding components of SRL should also improve, leading to an age-related increase in the correlation between EF and SRL. Two self-report instruments were used: the strategies for self-regulated learning survey (SSRLS) and the behavioural rating instrument of executive function (BRIEF). Strong correlations between the measures of EF and SRL were found, especially in areas associated with metacognitive processes. Correlations between EF and SRL were found, with weaker correlations between behavioural regulation and SRL were found to be weaker for the younger participants in the sample while the relationship between EF and SRL appears to grow stronger during the initial years of high school even though self-reported levels of EF along with motivation for SRL and important components of SRL such as goal setting and planning were found to decrease with age. Decreasing levels of motivation for learning during adolescence are speculated to moderate the deployment of SRL and EF in a school context.

Self-Regulated Learning : key strategies and their sources in a sample of adolescent males

This study identified the key self-regulated learning (SRL) strategies and their sources for nine school-aged adolescent males aged 15 to 17 years. The Self-Regulated Learning Interview Schedule (SRLIS) was used along with semi-structured interviews with the participants and their parents to elicit information on SRL strategies and contexts for the formation of self-regulatory habits. Early habit-forming experiences of the family home in relation to homework and study routines were found to form an important base for effective SRL. Teachers were identified as the most common source of SRL strategies with important formative experiences occurring during the first two years of high school.

Early school-based parent involvement, children’s self-regulated learning and academic achievement: An Australian longitudinal study

Parent involvement is widely accepted as being associated with children’s improved educational outcomes. However, the role of early school-based parent involvement is still being established. This study investigated the mediating role of self-regulated learning behaviors in the relationship between early school-based parent involvement and children’s academic achievement, using data from the Longitudinal Study of Australian Children (N = 2616). Family socioeconomic position, Aboriginal and Torres Strait Islander status, language background, child gender and cognitive competence, were controlled, as well home and community based parent involvement activity in order to derive a more confident interpretation of the results. Structural equation modeling analyses showed that children’s self-regulated learning behaviors fully mediated the relationships between school-based parent involvement at Grade 1 and children’s reading achievement at Grade 3. Importantly, these relationships were evident for children across all socio-economic backgrounds. Although there was no direct relationship between parent involvement at Grade 1 and numeracy achievement at Grade 3, parent involvement was indirectly associated with higher children’s numeracy achievement through children’s self-regulation of learning behaviors, though this relationship was stronger for children from middle and higher socio-economic backgrounds. Implications for policy and practice are discussed, and further research recommended.

Semi-supervised learning of WLAN radio maps

In this thesis a manifold learning method is applied to the problem of WLAN positioning and automatic radio map creation. Due to the nature of WLAN signal strength measurements, a signal map created from raw measurements results in non-linear distance relations between measurement points. These signal strength vectors reside in a high-dimensioned coordinate system. With the help of the so called Isomap-algorithm the dimensionality of this map can be reduced, and thus more easily processed. By embedding position-labeled strategic key points, we can automatically adjust the mapping to match the surveyed environment. The environment is thus learned in a semi-supervised way; gathering training points and embedding them in a two-dimensional manifold gives us a rough mapping of the measured environment. After a calibration phase, where the labeled key points in the training data are used to associate coordinates in the manifold representation with geographical locations, we can perform positioning using the adjusted map. This can be achieved through a traditional supervised learning process, which in our case is a simple nearest neighbors matching of a sampled signal strength vector. We deployed this system in two locations in the Kumpula campus in Helsinki, Finland. Results indicate that positioning based on the learned radio map can achieve good accuracy, especially in hallways or other areas in the environment where the WLAN signal is constrained by obstacles such as walls.