A Probabilistic Re-Intepretation of Confidence Scores in Multi-Exit Models

In this paper, we propose a new approach to train a deep neural network with multiple intermediate auxiliary classifiers, branching from it. These ‘multi-exits’ models can be used to reduce the inference time by performing early exit on the intermediate branches, if the confidence of the prediction is higher than a threshold. They rely on the assumption that not all the samples require the same amount of processing to yield a good prediction. In this paper, we propose a way to train jointly all the branches of a multi-exit model without hyper-parameters, by weighting the predictions from each branch with a trained confidence score. Each confidence score is an approximation of the real one produced by the branch, and it is calculated and regularized while training the rest of the model. We evaluate our proposal on a set of image classification benchmarks, using different neural models and early-exit stopping criteria.

Evaluating Elimination Thresholds and Stopping Criteria for Interventions Against the Vector-borne Macroparasitic Disease, Lymphatic Filariasis, Using Mathematical Modelling

We leverage the ability of the EPIFIL transmission model fit to field data for allowing calculations of the probabilities of transmission elimination and recrudescence once infection levels are predicted to fall below the threshold used in the WHO Transmission Assessment Surveys (TAS) versus site-specific model-estimated thresholds to evaluate the implications of using these thresholds for making Lymphatic Filariasis (LF) intervention stopping decisions. Our results, overall, indicate that understanding the underlying parasite transmission and extinction dynamics will be crucial for choosing the right intervention stopping thresholds, and indeed the right interventions connected with these thresholds, if we are to bring about the sustainable elimination of LF. They also warn that applying stopping criteria set for operational purposes without a full consideration of population dynamics, as employed in the current TAS strategy, could, by risking infection recrudescence especially over the long-term, seriously undermine the goal of achieving global LF elimination.

The effects of the cross-entropy stopping criterion and quadrature amplitude modulation on iterative turbo decoding performance

One of the most often-used stopping criteria is the cross-entropy stopping criterion (CESC). The CESC can stop turbo decoder iterations early by calculating mutual information improvements while maintaining bit error rate (BER) performance. Most research on iterative turbo decoding stopping criteria has utilised low-modulation methods, such as binary phase-shift keying. However, a high-speed network requires high modulation to transfer data at high speeds. Hence, a high modulation technique needs to be integrated into the CESC to match its speed. Therefore, the present paper investigated and analysed the effects of the CESC and quadrature amplitude modulation (QAM) on iterative turbo decoding. Three thresholds were simulated and tested under four situations: different code rates, different QAM formats, different code generators, and different frame sizes. The results revealed that in most situations, the use of CESC is suitable only when the signal-to-noise ratio (SNR) is high. This is because the CESC significantly reduces the average iteration number (AIN) while maintaining the BER. The CESC can terminate early at a high SNR and save more than 40% AIN compared with the fixed stopping criterion. Meanwhile, at a low SNR, the CESC fails to terminate early, which results in maximum AIN.

HBeAg-positive patients with HBsAg < 100 IU/mL and negative HBV RNA have lower risk of virological relapse after nucleos(t)ide analogues cessation

Background Nucleos(t)ide analogues (NAs) cessation is not widely practiced and remains a controversial, but highly relevant subject in patients infected with hepatitis B virus (HBV). We aimed to explore the related factors for safe NAs cessation. Methods This is a multicenter prospective cohort study. Overall, 139 initially HBV e antigen (HBeAg)-positive patients meeting the stopping criteria were included in 12 hospitals in China. Enrolled patients ceased NAs and were followed up every 3 months for 24 months or until clinical relapse (CR). Results The 24 month cumulative rates of virological relapse (VR), CR, HBeAg reversion and HBV surface antigen (HBsAg) loss were 50.4, 24.5, 11.5 and 9.4%, respectively. Patients with end of treatment (EOT) HBsAg < 100 IU/mL plus negative HBV RNA had the lowest 24 month cumulative VR rate (5 vs 58%, p < 0.001). EOT HBsAg ≥ 2 log10 IU/mL [odds ratio (OR) = 6.686, p = 0.006], EOT positive HBV RNA (OR = 3.453, p = 0.008) and EOT hepatitis B core-related antigen (HBcrAg) ≥ 4log U/mL (OR = 3.702, p = 0.002) were found to independently predict the risk of VR. To predict VR, the area under the receiver-operating characteristic (AUROC) value of the EOT HBsAg < 100 IU/mL plus EOT HBV RNA negative was 0.698 (p < 0.001), which was higher than other parameters alone or combinations. Conclusions NAs cessation is suitable only for a small and selected patients. An EOT HBsAg < 100 IU/mL and EOT negative HBV RNA identified a patient with low risk of off-treatment VR.

Generating Set Search Method with Optimal Bind Tolerance for Solving the Economic Load Dispatch Problem of Thermal Stations

A non-derivative direct search approach called Generating Set Search (GSS) algorithm with varying bind tolerance to solve non-convex Economic Load Dispatch problem of the thermal stations in Nigeria is presented. A complete poll was carried out with initial mesh size of 1.0, expansion factor of 2.0 and contraction factor of 0.5. The binding tolerance was varied from 100 – 2200 with an increment of 100. The stopping criteria were based on the following: mesh tolerance of 0.000001, maximum iteration of 1500 and maximum function evaluation of 30000. The Economic Load Dispatch of 2500 MW, 3000 MW, 3500 MW and 4000 MW produced optimal solutions at binding tolerances of 500, 600, 1100, and 1600 respectively. The economic cost (measured in quantity of fuel) obtained for the dispatch of 2500 MW, 3000 MW, 3500 MW and 4000 MW were 83577.6936190168 MMBTU/hr, 83577.6936667599 MMBTU/hr, 83577.6937160183 MMBTU/hr and 83577.694264612 MMBTU/hr respectively. The evaluations carried out on the function in order to obtain the best solution were 1484, 5709, 6895 and 7556 for 2500 MW, 3000 MW, 3500 MW and 4000 MW of load respectively. Although the optimal bind tolerances had more iterations and evaluations, these can be traded off for the best solutions offered.<br>

Generating Set Search Method with Optimal Bind Tolerance for Solving the Economic Load Dispatch Problem of Thermal Stations

A non-derivative direct search approach called Generating Set Search (GSS) algorithm with varying bind tolerance to solve non-convex Economic Load Dispatch problem of the thermal stations in Nigeria is presented. A complete poll was carried out with initial mesh size of 1.0, expansion factor of 2.0 and contraction factor of 0.5. The binding tolerance was varied from 100 – 2200 with an increment of 100. The stopping criteria were based on the following: mesh tolerance of 0.000001, maximum iteration of 1500 and maximum function evaluation of 30000. The Economic Load Dispatch of 2500 MW, 3000 MW, 3500 MW and 4000 MW produced optimal solutions at binding tolerances of 500, 600, 1100, and 1600 respectively. The economic cost (measured in quantity of fuel) obtained for the dispatch of 2500 MW, 3000 MW, 3500 MW and 4000 MW were 83577.6936190168 MMBTU/hr, 83577.6936667599 MMBTU/hr, 83577.6937160183 MMBTU/hr and 83577.694264612 MMBTU/hr respectively. The evaluations carried out on the function in order to obtain the best solution were 1484, 5709, 6895 and 7556 for 2500 MW, 3000 MW, 3500 MW and 4000 MW of load respectively. Although the optimal bind tolerances had more iterations and evaluations, these can be traded off for the best solutions offered.<br>

Stopping criteria for ending autonomous, single detector radiological source searches

While the localization of radiological sources has traditionally been handled with statistical algorithms, such a task can be augmented with advanced machine learning methodologies. The combination of deep and reinforcement learning has provided learning-based navigation to autonomous, single-detector, mobile systems. However, these approaches lacked the capacity to terminate a surveying/search task without outside influence of an operator or perfect knowledge of source location (defeating the purpose of such a system). Two stopping criteria are investigated in this work for a machine learning navigated system: one based upon Bayesian and maximum likelihood estimation (MLE) strategies commonly used in source localization, and a second providing the navigational machine learning network with a “stop search” action. A convolutional neural network was trained via reinforcement learning in a 10 m × 10 m simulated environment to navigate a randomly placed detector-agent to a randomly placed source of varied strength (stopping with perfect knowledge during training). The network agent could move in one of four directions (up, down, left, right) after taking a 1 s count measurement at the current location. During testing, the stopping criteria for this navigational algorithm was based upon a Bayesian likelihood estimation technique of source presence, updating this likelihood after each step, and terminating once the confidence of the source being in a single location exceeded 0.9. A second network was trained and tested with similar architecture as the previous but which contained a fifth action: for self-stopping. The accuracy and speed of localization with set detector and source initializations were compared over 50 trials of MLE-Bayesian approach and 1000 trials of the CNN with self-stopping. The statistical stopping condition yielded a median localization error of ~1.41 m and median localization speed of 12 steps. The machine learning stopping condition yielded a median localization error of 0 m and median localization speed of 17 steps. This work demonstrated two stopping criteria available to a machine learning guided, source localization system.