Learning from falling.

Walkers fall frequently, especially during infancy. Children (15-, 21-, 27-, 33-, and 39-month-olds) and adults were tested in a novel foam pit paradigm to examine age-related changes in the relationship between falling and prospective control of locomotion. In trial 1, participants walked and fell into a deformable foam pit marked with distinct visual cues. Although children in all 5 age groups required multiple trials to learn to avoid falling, the number of children who showed adult-like, 1-trial learning increased with age. Exploration and alternative locomotor strategies increased dramatically on learning criterion trials and displays of negative affect were limited. Learning from falling is discussed in terms of the immediate and long-term effects of falling on prospective control of locomotion.

Specialized motor-driven dusp1 expression in the song systems of multiple lineages of vocal learning birds.

Mechanisms for the evolution of convergent behavioral traits are largely unknown. Vocal learning is one such trait that evolved multiple times and is necessary in humans for the acquisition of spoken language. Among birds, vocal learning is evolved in songbirds, parrots, and hummingbirds. Each time similar forebrain song nuclei specialized for vocal learning and production have evolved. This finding led to the hypothesis that the behavioral and neuroanatomical convergences for vocal learning could be associated with molecular convergence. We previously found that the neural activity-induced gene dual specificity phosphatase 1 (dusp1) was up-regulated in non-vocal circuits, specifically in sensory-input neurons of the thalamus and telencephalon; however, dusp1 was not up-regulated in higher order sensory neurons or motor circuits. Here we show that song motor nuclei are an exception to this pattern. The song nuclei of species from all known vocal learning avian lineages showed motor-driven up-regulation of dusp1 expression induced by singing. There was no detectable motor-driven dusp1 expression throughout the rest of the forebrain after non-vocal motor performance. This pattern contrasts with expression of the commonly studied activity-induced gene egr1, which shows motor-driven expression in song nuclei induced by singing, but also motor-driven expression in adjacent brain regions after non-vocal motor behaviors. In the vocal non-learning avian species, we found no detectable vocalizing-driven dusp1 expression in the forebrain. These findings suggest that independent evolutions of neural systems for vocal learning were accompanied by selection for specialized motor-driven expression of the dusp1 gene in those circuits. This specialized expression of dusp1 could potentially lead to differential regulation of dusp1-modulated molecular cascades in vocal learning circuits.

Improvement in visual search with practice: mapping learning-related changes in neurocognitive stages of processing.

Practice can improve performance on visual search tasks; the neural mechanisms underlying such improvements, however, are not clear. Response time typically shortens with practice, but which components of the stimulus-response processing chain facilitate this behavioral change? Improved search performance could result from enhancements in various cognitive processing stages, including (1) sensory processing, (2) attentional allocation, (3) target discrimination, (4) motor-response preparation, and/or (5) response execution. We measured event-related potentials (ERPs) as human participants completed a five-day visual-search protocol in which they reported the orientation of a color popout target within an array of ellipses. We assessed changes in behavioral performance and in ERP components associated with various stages of processing. After practice, response time decreased in all participants (while accuracy remained consistent), and electrophysiological measures revealed modulation of several ERP components. First, amplitudes of the early sensory-evoked N1 component at 150 ms increased bilaterally, indicating enhanced visual sensory processing of the array. Second, the negative-polarity posterior-contralateral component (N2pc, 170-250 ms) was earlier and larger, demonstrating enhanced attentional orienting. Third, the amplitude of the sustained posterior contralateral negativity component (SPCN, 300-400 ms) decreased, indicating facilitated target discrimination. Finally, faster motor-response preparation and execution were observed after practice, as indicated by latency changes in both the stimulus-locked and response-locked lateralized readiness potentials (LRPs). These electrophysiological results delineate the functional plasticity in key mechanisms underlying visual search with high temporal resolution and illustrate how practice influences various cognitive and neural processing stages leading to enhanced behavioral performance.

A comparison of the effectiveness of the team-based learning readiness assessments completed at home to those completed in class.

PURPOSE: The readiness assurance process (RAP) of team-based learning (TBL) is an important element that ensures that students come prepared to learn. However, the RAP can use a significant amount of class time which could otherwise be used for application exercises. The authors administered the TBL-associated RAP in class or individual readiness assurance tests (iRATs) at home to compare medical student performance and learning preference for physiology content. METHODS: Using cross-over study design, the first year medical student TBL teams were divided into two groups. One group was administered iRATs and group readiness assurance tests (gRATs) consisting of physiology questions during scheduled class time. The other group was administered the same iRAT questions at home, and did not complete a gRAT. To compare effectiveness of the two administration methods, both groups completed the same 12-question physiology assessment during dedicated class time. Four weeks later, the entire process was repeated, with each group administered the RAP using the opposite method. RESULTS: The performance on the physiology assessment after at-home administration of the iRAT was equivalent to performance after traditional in-class administration of the RAP. In addition, a majority of students preferred the at-home method of administration and reported that the at-home method was more effective in helping them learn course content. CONCLUSION: The at-home administration of the iRAT proved effective. The at-home administration method is a promising alternative to conventional iRATs and gRATs with the goal of preserving valuable in-class time for TBL application exercises.

Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

Learning and Exploiting Camera Geometry for Computer Vision

This work explores the use of statistical methods in describing and estimating camera poses, as well as the information feedback loop between camera pose and object detection. Surging development in robotics and computer vision has pushed the need for algorithms that infer, understand, and utilize information about the position and orientation of the sensor platforms when observing and/or interacting with their environment.

The first contribution of this thesis is the development of a set of statistical tools for representing and estimating the uncertainty in object poses. A distribution for representing the joint uncertainty over multiple object positions and orientations is described, called the mirrored normal-Bingham distribution. This distribution generalizes both the normal distribution in Euclidean space, and the Bingham distribution on the unit hypersphere. It is shown to inherit many of the convenient properties of these special cases: it is the maximum-entropy distribution with fixed second moment, and there is a generalized Laplace approximation whose result is the mirrored normal-Bingham distribution. This distribution and approximation method are demonstrated by deriving the analytical approximation to the wrapped-normal distribution. Further, it is shown how these tools can be used to represent the uncertainty in the result of a bundle adjustment problem.

Another application of these methods is illustrated as part of a novel camera pose estimation algorithm based on object detections. The autocalibration task is formulated as a bundle adjustment problem using prior distributions over the 3D points to enforce the objects' structure and their relationship with the scene geometry. This framework is very flexible and enables the use of off-the-shelf computational tools to solve specialized autocalibration problems. Its performance is evaluated using a pedestrian detector to provide head and foot location observations, and it proves much faster and potentially more accurate than existing methods.

Finally, the information feedback loop between object detection and camera pose estimation is closed by utilizing camera pose information to improve object detection in scenarios with significant perspective warping. Methods are presented that allow the inverse perspective mapping traditionally applied to images to be applied instead to features computed from those images. For the special case of HOG-like features, which are used by many modern object detection systems, these methods are shown to provide substantial performance benefits over unadapted detectors while achieving real-time frame rates, orders of magnitude faster than comparable image warping methods.

The statistical tools and algorithms presented here are especially promising for mobile cameras, providing the ability to autocalibrate and adapt to the camera pose in real time. In addition, these methods have wide-ranging potential applications in diverse areas of computer vision, robotics, and imaging.