Improved Model Configuration Strategies for Kannada Handwritten Numeral Recognition

Handwritten numeral recognition has been an important area in the domain of pattern classification. The task becomes even more daunting when working with non-Roman numerals. While convolutional neural networks are the preferred choice for modeling the image data, the conception of techniques to obtain faster convergence and accurate results still poses an enigma to the researchers. In this paper, we present new methods for the initialization and the optimization of the traditional convolutional neural network architecture to obtain better results for Kannada numeral images. Specifically, we propose two different methods- an encoderdecoder setup for unsupervised training and weight initialization, and a particle swarm optimization strategy for choosing the ideal architecture configuration of the CNN. Unsupervised initial training of the architecture helps for a faster convergence owing to more task-suited weights as compared to random initialization while the optimization strategy is helpful to reduce the time required for the manual iterative approach of architecture selection. The proposed setup is trained on varying handwritten Kannada numerals. The proposed approaches are evaluated on two different datasets: a standard Dig-MNIST dataset and a custom-built dataset. Significant improvements across multiple performance metrics are observed in our proposed system over the traditional CNN training setup. The improvement in results makes a strong case for relying on such methods for faster and more accurate training and inference of digit classification, especially when working in the absence of transfer learning.

Feature Fusion-Based Siamese Region Proposal Network for Ultrasound Tracking

Object tracking based on ultrasound image navigation can effectively reduce damage to healthy tissues in radiotherapy. In this study, we propose a deep Siamese network based on feature fusion. Whilst adopting MobileNetV2 as the backbone, an unsupervised training strategy is introduced to enrich the volume of the samples. The region proposal network module is designed to predict the location of the target, and a non-maximum suppression-based postprocessing algorithm is designed to refine the tracking results. Moreover, the proposed method is evaluated in the Challenge on Liver Ultrasound Tracking dataset and the self-collected dataset, which proves the need for the improvement and the effectiveness of the algorithm.

Mode-assisted joint training of deep Boltzmann machines

The deep extension of the restricted Boltzmann machine (RBM), known as the deep Boltzmann machine (DBM), is an expressive family of machine learning models which can serve as compact representations of complex probability distributions. However, jointly training DBMs in the unsupervised setting has proven to be a formidable task. A recent technique we have proposed, called mode-assisted training, has shown great success in improving the unsupervised training of RBMs. Here, we show that the performance gains of the mode-assisted training are even more dramatic for DBMs. In fact, DBMs jointly trained with the mode-assisted algorithm can represent the same data set with orders of magnitude lower number of total parameters compared to state-of-the-art training procedures and even with respect to RBMs, provided a fan-in network topology is also introduced. This substantial saving in number of parameters makes this training method very appealing also for hardware implementations.

Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT

Due to major artifacts and uncalibrated Hounsfield units (HU), cone-beam computed tomography (CBCT) cannot be used readily for diagnostics and therapy planning purposes. This study addresses image-to-image translation by convolutional neural networks (CNNs) to convert CBCT to CT-like scans, comparing supervised to unsupervised training techniques, exploiting a pelvic CT/CBCT publicly available dataset. Interestingly, quantitative results were in favor of supervised against unsupervised approach showing improvements in the HU accuracy (62% vs. 50%), structural similarity index (2.5% vs. 1.1%) and peak signal-to-noise ratio (15% vs. 8%). Qualitative results conversely showcased higher anatomical artifacts in the synthetic CBCT generated by the supervised techniques. This was motivated by the higher sensitivity of the supervised training technique to the pixel-wise correspondence contained in the loss function. The unsupervised technique does not require correspondence and mitigates this drawback as it combines adversarial, cycle consistency, and identity loss functions. Overall, two main impacts qualify the paper: (a) the feasibility of CNN to generate accurate synthetic CT from CBCT images, which is fast and easy to use compared to traditional techniques applied in clinics; (b) the proposal of guidelines to drive the selection of the better training technique, which can be shifted to more general image-to-image translation.

Evaluation Algorithm of Teaching Work Quality in Colleges and Universities Based on Deep Denoising Autoencoder Network

One of the most significant components of the teaching department’s evaluation of teaching quality is evaluating teachers’ performance. With the acceleration of educational informatization, modern information processing technology can be used effectively to evaluate teachers’ teaching quality in traditional teaching. In this context, combined with some computational intelligence algorithms, it is critical to developing a targeted teaching quality evaluation system. This paper studies teacher teaching evaluation’s characteristics and existing problems and analyzes the fundamental theories and methods of teacher teaching evaluation in colleges and universities. A novel combination of deep denoising autoencoder and support vector machine was proposed for evaluating teacher’s teaching quality. Moreover, support vector regression is used to predict the model’s output layer to achieve supervised assessment prediction. To capture the data’s key properties, the model comprises numerous hidden layers and conducts various feature transformations during unsupervised training to minimize the mean square error between the reconstructed output data and the original input data. As a result, the proposed model achieved the highest recognition accuracy of 85.23% and convergence compared to other models. Thus, the method can be employed to evaluate and forecast the quality of university teaching activity successfully.

When unsupervised training benefits category learning

Humans continuously categorise inputs, but only rarely receive explicit feedback as to whether or not they are correct. This implies that they may be integrating unsupervised information together with their sparse supervised data -- a form of semi-supervised learning. However, experiments testing semi-supervised learning are rare, and are bedevilled with conflicting results about whether the unsupervised information affords any benefit. Here, we suggest that one important factor that has been paid insufficient attention is the alignment between subjects' internal representations of the stimulus material and the experimenter-defined representations that determine success in the tasks. Subjects' representations are shaped by prior biases and experience, and unsupervised learning can only be successful if the alignment suffices. Otherwise, unsupervised learning might harmfully strengthen incorrect assumptions. To test this hypothesis, we conducted an experiment in which subjects initially categorise items along a salient, but task-irrelevant, dimension, and only recover the correct categories when sufficient feedback draws their attention to the subtle, task-relevant, stimulus dimensions. By withdrawing feedback at different stages along this learning curve, we tested whether unsupervised learning improves or worsens performance when internal stimulus representations and task are sufficiently or insufficiently aligned, respectively. Our results demonstrate that unsupervised learning can indeed have opposing effects on subjects' learning. We also discuss factors limiting the degree to which such effects can be predicted from momentary performance. Our work implies that predicting and understanding human category learning in particular tasks requires assessment and consideration of the representational spaces that subjects entertain for the materials involved in those tasks. These considerations not only apply to studies in the lab, but could also help improve the design of tutoring systems and instruction.

Brief Vigorous Stair Climbing Effectively Improves Cardiorespiratory Fitness in Patients With Coronary Artery Disease: A Randomized Trial

Background: Cardiac rehabilitation exercise reduces the risk of secondary cardiovascular disease. Interval training is a time-efficient alternative to traditional cardiac rehabilitation exercise and stair climbing is an accessible means. We aimed to assess the effectiveness of a high-intensity interval stair climbing intervention on improving cardiorespiratory fitness (V˙O2peak) compared to standard cardiac rehabilitation care.Methods: Twenty participants with coronary artery disease (61 ± 7 years, 18 males, two females) were randomly assigned to either traditional moderate-intensity exercise (TRAD) or high-intensity interval stair climbing (STAIR). V˙O2peak was assessed at baseline, following 4 weeks of six supervised exercise sessions and after 8 weeks of ~24 unsupervised exercise sessions. TRAD involved a minimum of 30 min at 60–80%HRpeak, and STAIR consisted of three bouts of six flights of 12 stairs at a self-selected vigorous intensity (~90 s/bout) separated by recovery periods of walking (~90 s). This study was registered as a clinical trial at clinicaltrials.gov (NCT03235674).Results: Two participants could not complete the trial due to the time commitment of the testing visits, leaving n = 9 in each group who completed the interventions without any adverse events. V˙O2peak increased after supervised and unsupervised training in comparison to baseline for both TRAD [baseline: 22.9 ± 2.5, 4 weeks (supervised): 25.3 ± 4.4, and 12 weeks (unsupervised): 26.5 ± 4.8 mL/kg/min] and STAIR [baseline: 21.4 ± 4.5, 4 weeks (supervised): 23.4 ± 5.6, and 12 weeks (unsupervised): 25 ± 6.2 mL/kg/min; p (time) = 0.03]. During the first 4 weeks of training (supervised) the STAIR vs. TRAD group had a higher %HRpeak (101 ± 1 vs. 89 ± 1%; p ≤ 0.001), across a shorter total exercise time (7.1 ± 0.1 vs. 36.7 ± 1.1 min; p = 0.009). During the subsequent 8 weeks of unsupervised training, %HRpeak was not different (87 ± 8 vs. 96 ± 8%; p = 0.055, mean ± SD) between groups, however, the STAIR group continued to exercise for less time per session (10.0 ± 3.2 vs. 24.2 ± 17.0 min; p = 0.036).Conclusions: Both brief, vigorous stair climbing, and traditional moderate-intensity exercise are effective in increasing V˙O2peak, in cardiac rehabilitation exercise programmes.