Generative Learning Model Research Articles

Learning Generative Models of Scene Features

We present a method for learning a set of generative models which are suitable for representing selected image-domain features of a scene as a function of changes in the camera viewpoint. Such models are important for robotic tasks, such as probabilistic position estimation (i.e. localization), as well as visualization. Our approach entails the automatic selection of the features, as well as the synthesis of models of their visual behavior. The model we propose is capable of generating maximum-likelihood views, as well as a measure of the likelihood of a particular view from a particular camera position. Training the models involves regularizing observations of the features from known camera locations. The uncertainty of the model is evaluated using cross validation, which allows for a priori evaluation of features and their attributes. The features themselves are initially selected as salient points by a measure of visual attention, and are tracked across multiple views. While the motivation for this work is for robot localization, the results have implications for image interpolation, image-based scene reconstruction and object recognition. This paper presents a formulation of the problem and illustrative experimental results.

Learning generative models of natural images

This work proposes an unsupervised learning process for analysis of natural images. The derivation is based on a generative model, a stochastic coin-flip process directly operating on many disjoint multivariate Gaussian distributions. Following the maximal likelihood principle and using the Potts encoding, the goodness-of-fit of the generative model to tremendous patches randomly sampled from natural images is quantitatively expressed by an objective function subject to a set of constraints. By further combination of the objective function and the minimal wiring criterion, we achieve a mixed integer and linear programming. A hybrid of the mean field annealing and the gradient descent method is applied to the mathematical framework and produces three sets of interactive dynamics for the learning process. Numerical simulations show that the learning process is effective for extraction of orientation, localization and bandpass features and the generative model can make an ensemble of a sparse code for natural images.

Weaving Chains of Affect and Cognition: A Young Child's Understanding of CD-ROM Talking Books

This qualitative case study employed Wittrock's Generative Learning Model to examine in-depth one kindergarten child's comprehension when reading considerate and inconsiderate CD-ROM talking books in a classroom computer center. A CD-ROM talking book consists of a story told through multimedia modes of information that has been digitized on a CD. Considerate CD-ROM talking books are those that include multimedia effects that are congruent with and integral to the story. Inconsiderate CD-ROM talking books are those that include multimedia effects that are incongruent with or incidental to the story. Findings indicate that considerate CD-ROM talking books supported the child's understanding and retelling of the story and involved meaning-making processes that wove together affective responses, cognitive processes, and metacognitive activity; however, inconsiderate CD-ROM talking books resulted in the child's inability to retell the story in a cohesive way and fostered passive viewing. Implications for research and practice are drawn.

Determining Generative Models of Objects Under Varying Illumination: Shape and Albedo from Multiple Images Using SVDand Integrability

We describe a method of learning generative models of objects from a set of images of the object under different, and unknown, illumination. Such a model allows us to approximate the objects‘ appearance under a range of lighting conditions. This work is closely related to photometric stereo with unknown light sources and, in particular, to the use of Singular Value Decomposition (SVD) to estimate shape and albedo from multiple images up to a linear transformation (Hayakawa, 1994). Firstly we analyze and extend the SVD approach to this problem. We demonstrate that it applies to objects for which the dominant imaging effects are Lambertian reflectance with a distant light source and a background ambient term. To determine that this is a reasonable approximation we calculate the eigenvectors of the SVD on a set of real objects, under varying lighting conditions, and demonstrate that the first few eigenvectors account for most of the data in agreement with our predictions. We then analyze the linear ambiguities in the SVD approach and demonstrate that previous methods proposed to resolve them (Hayakawa, 1994) are only valid under certain conditions. We discuss alternative possibilities and, in particular, demonstrate that knowledge of the object class is sufficient to resolve this problem. Secondly, we describe the use of surface consistency for putting constraints on the possible solutions. We prove that this constraint reduces the ambiguities to a subspace called the generalized bas relief ambiguity (GBR) which is inherent in the Lambertian reflectance function (and which can be shown to exist even if attached and cast shadows are present (Belhumeur et al., 1997)). We demonstrate the use of surface consistency to solve for the shape and albedo up to a GBR and describe, and implement, a variety of additional assumptions to resolve the GBR. Thirdly, we demonstrate an iterative algorithm that can detect and remove some attached shadows from the objects thereby increasing the accuracy of the reconstructed shape and albedo.

Understanding a generative learning model of instruction: A case study of elementary teacher planning

(1996). Understanding a generative learning model of instruction: A case study of elementary teacher planning. Journal of Science Teacher Education: Vol. 7, No. 2, pp. 95-122.

Gender Difference, Misconceptions and Instruction in Science

Abstract The study examined into the relationship between gender and students' misconceptions in science. Two different groups were treated with two different teaching strategies, namely, teaching strategy 1, which is basically didactic in nature, and teaching strategy 11, which incorporates students' misconceptions and applies the Generative Learning Model. Two groups of secondary three students (N=26,27; randomly sampled), underwent 6 weeks of instruction, with the respective strategies mentioned above. Each group consisted of male and female students, the numbers of which resulted from the grouping based on their academic achievements. A constructed and validated diagnostic instrument was used as a means to measure the effectiveness of these two teaching strategies. The findings showed that gender differences did not relate well to students' misconceptions in science. The implications of this finding are discussed.

Un modèle d'enseignement génératif de la photosynthèse

A generative learning strategy about photosynthesis , Miles Barker, Malcolm Carr Three strategies for teaching and learning about photosynthesis are described and criticized on the grounds that none of them promotes understanding of photosynthesis as a carbohydrate-producing process in a way which can be related to students’prior knowledge. A new strategy exploring the material aspects of photosynthesis (carbohydrate production) based on the generative learning model of Osborne and Wittrock (1985) has been developed. A teaching package entitled “Where Does The Wood Come From ?" has been trialled with a class of 26 fouth formers (14-year-olds). Seventy-one per cent of the students acquired a view of photosynthesis as a carbohydrate-producing process. Some novel techniques for implementing constructivist theory in the classroom (investigations, surveys, a self-teach booklet, check-points) are described. Modifications to the generative learning model itself, especially its apparently sequential nature, are suggested.

Introducing technological applications into the physics classroom: help or hindrance for learning?

In the past school physics has tended to emphasize the philosophical, experimental and mathematical aspects of the subject. This paper explores a teaching sequence, based on some of the ideas of the generative learning model, to introduce technological applications into the physics classroom (17‐18‐year‐old students). The development and evaluation of the teaching sequence, about electrical capacitance, is reported. The evaluation consisted of three interlinked strategies: classroom observation and student interviews, teacher interviews, and an attitudinal scale used in conjunction with student interviews. Although this was a relatively small scale experiment, the students generally responded positively both to the introduction of technological applications and to the learning environment.

Establishing an epistemological base for science teaching in the light of contemporary notions of the nature of science and of how children learn science

AbstractThe quality of science education is a pervasive concern in educational improvement efforts. This article examines two aspects of science education that have been identified as being of concern: Is the material presented in science classrooms representative of science “as it really is” and, secondly, what is the bearing of notions of “children's science” on curriculum reform? A review of the literature in both areas shows that educational theory has, so far, had little effect on classroom practice. An attempt is made to synthesize contemporary views on the tentative nature of science with a generative model of learning in science in order to establish a common epistemological base which could be used as a template for curriculum reform efforts.

Teaching technological applications in the physics classroom

A teaching unit, designed to improve links between physics concepts and students' existing know ledge, has been developed and evaluated. Based on key propositions of the generative learning model, attention gaining activities were introduced to facilitate new ideas being linked to relevant frameworks already held by the learner. Evaluation suggests that introducing an appropriate technological focus, combining activities and using a suitable teaching sequence generally enhance the learning of senior secondary school physics. Student perceptions of the teaching unit were probably enhanced in part because the approach used was new and different. However, when a second teaching unit, based on the Doppler effect, was introduced to the class about four weeks later, very similar results were obtained with the same classes and other classes. The capacitance unit was also used in other schools with similar positive results from both teachers and students.

Teaching and learning about photosynthesis. Part 2: A generative learning strategy

A new strategy exploring the material aspects of photosynthesis (carbohydrate production) based on the generative learning model of Osborne and Wittrock (1985) has been developed. A teaching package entitled ‘Where Does The Wood Come From?’ has been trialled by an experienced and sympathetic teacher with a middle ability class of 26 fourth formers (14‐year‐olds). Seventy‐one per cent of the students acquired a view of photosynthesis as a carbohydrate‐producing process. This contrasts with the usual guided discovery strategy, where a food‐making view is the major outcome. Some novel techniques for implementing constructivist theory in the classroom (investigations, surveys, a self‐teach booklet, checkpoints) are described. Modifications to the generative learning model itself, especially its apparently sequential nature, are suggested.

The Effects of Participatory Expectations in Classroom Discussion on the Learning of Science Vocabulary

To study why discussion appears to improve vocabulary learning, children in two intact classes were divided into three subgroups and taught science vocabulary on 3 successive days. For each lesson, one subgroup was told they would not be called on and were to learn by listening. Of the other two subgroups, only one was actually called on (Called-on) and the other merely anticipated being called on (Ignored). Two additional classes served as comparison groups. Students in the treatment group generally outperformed those in the two comparison groups on the posttests, regardless of condition. Within the treatment group, students in both anticipatory conditions (Called-on and Ignored) significantly outperformed students in the Listening condition on the delayed posttest, but not on the immediate posttest. These results are discussed within the context of a generative model of vocabulary learning.

The Generative Learning Model and its Implications for Science Education

(1985). The Generative Learning Model and its Implications for Science Education. Studies in Science Education: Vol. 12, No. 1, pp. 59-87.

Reduced Fidelity Motor Skill Training

The study investigated if training using text and pictures (reduced fidelity) would lead to better performance of a motor skill task. The motor skill was a two-hand tracking task similar to a video game. Task performance was assessed by two measures: accuracy and strategy. Using the generative learning model and the characteristics of the analytic-holistic cognitive style, four instructional treatments and a control condition were developed. Two instructional treatments presented training in a holistic mode and two treatments used an analytic mode. Within each mode, one treatment presented generative instructions. One hundred twenty-four undergraduate college students were randomly assigned to treatments. Results included that generative processing lead to more accurate performance. Strategy was influenced by interactions among treatment, cognitive style, and sex of the learner.

Fieldwork in Geography and Long-Term Memory Structures

This study was based on the model of memory proposed by Gagne and White (1978). Three treatments were compared in their effects on eighth and ninth graders’ learning and retention of geographical facts and skills. In addition to a basic program, the first treatment had an active excursion, the second had a passive excursion, and the third had no excursion. The excursion treatments were slightly better than the control on initial learning. In retention, the active excursion treatment showed marked superiority over the others, which supports the Gagné and White model and Wittrock’s (1974) generative model of learning, and which has implications for the management of excursions.

Fieldwork in Geography and Long-term Memory Structures

This study was based on the model of memory proposed by Gagne and White (1978). Three treatments were compared in their effects on eighth and ninth graders’ learning and retention of geographical facts and skills. In addition to a basic program, the first treatment had an active excursion, the second had a passive excursion, and the third had no excursion. The excursion treatments were slightly better than the control on initial learning. In retention, the active excursion treatment showed marked superiority over the others, which supports the Gagné and White model and Wittrock’s (1974) generative model of learning, and which has implications for the management of excursions.

How Do Elementary School Teachers Plan? The Nature of Planning and Influences on It

The Elementary School Journal Volume 81, Number 1 ? 1980 by The University of Chicago 0013-5984/81/8101-0008$0o1.00 In colleges and universities across the country, prospective and practicing teachers and administrators take courses on planning-curriculum-planning and instructional planning. During the courses, teachers and administrators usually study an objectives-first model of curriculumplanning. This model has four steps. Planners are expected to 1. Formulate objectives 2. Choose appropriate learning activities

A Generative Model of Mathematics Learning

The learning of mathematics is presented as a cognitive process, rather than as a behavioristic one. A generative model of mathematics learning is described. Empirical research relevant to the model is discussed. Implications for teaching mathematics as a generative process are summarized.

Training in the small department : by James H. Auten Charles C. Thomas (301–327 East Lawrence Avenue, Springfield, Illinois 62703), 1973, 132 pp., hardcover—$5.75

Deep generative models seek to recover the process with which the observed data was generated. They may be used to synthesize new samples or to subsequently extract representations. Successful approaches in the domain of images are driven by several core inductive biases. However, a bias to account for the compositional way in which humans structure a visual scene in terms of objects has frequently been overlooked. In this work, we investigate object compositionality as an inductive bias for Generative Adversarial Networks (GANs). We present a minimal modification of a standard generator to incorporate this inductive bias and find that it reliably learns to generate images as compositions of objects. Using this general design as a backbone, we then propose two useful extensions to incorporate dependencies among objects and background. We extensively evaluate our approach on several multi-object image datasets and highlight the merits of incorporating structure for representation learning purposes. In particular, we find that our structured GANs are better at generating multi-object images that are more faithful to the reference distribution. More so, we demonstrate how, by leveraging the structure of the learned generative process, one can ‘invert’ the learned generative model to perform unsupervised instance segmentation. On the challenging CLEVR dataset, it is shown how our approach is able to improve over other recent purely unsupervised object-centric approaches to image generation.