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.

Revisiting Batch Normalization for Training Low-Latency Deep Spiking Neural Networks From Scratch

Spiking Neural Networks (SNNs) have recently emerged as an alternative to deep learning owing to sparse, asynchronous and binary event (or spike) driven processing, that can yield huge energy efficiency benefits on neuromorphic hardware. However, SNNs convey temporally-varying spike activation through time that is likely to induce a large variation of forward activation and backward gradients, resulting in unstable training. To address this training issue in SNNs, we revisit Batch Normalization (BN) and propose a temporal Batch Normalization Through Time (BNTT) technique. Different from previous BN techniques with SNNs, we find that varying the BN parameters at every time-step allows the model to learn the time-varying input distribution better. Specifically, our proposed BNTT decouples the parameters in a BNTT layer along the time axis to capture the temporal dynamics of spikes. We demonstrate BNTT on CIFAR-10, CIFAR-100, Tiny-ImageNet, event-driven DVS-CIFAR10 datasets, and Sequential MNIST and show near state-of-the-art performance. We conduct comprehensive analysis on the temporal characteristic of BNTT and showcase interesting benefits toward robustness against random and adversarial noise. Further, by monitoring the learnt parameters of BNTT, we find that we can do temporal early exit. That is, we can reduce the inference latency by ~5 − 20 time-steps from the original training latency. The code has been released at https://github.com/Intelligent-Computing-Lab-Yale/BNTT-Batch-Normalization-Through-Time.

Three Barriers to Effective Programs with Payment for Ecosystem Services: Behavioral Responses in a Computer-Based Experiment

Financial incentives in the form of payment for ecosystem services (PES) can encourage participation in voluntary conservation programs, but real-world experience with PES is limited for services such as the provision of endangered species habitats. A computer-based laboratory experiment with 139 US college students as subjects suggests there are three barriers to effective PES programs: (1) financial rewards can crowd out altruism—low-level PES in the experiment was less effective than the same program without PES; (2) landowners may assuage guilt over destroying habitats by making contributions to ineffective conservation programs—participants often paired destruction of habitat with token contributions to conservation efforts; and (3) landowners may strategically exit conservation agreements in ways that are detrimental to wildlife—a large proportion of participants chose to leave agreements and destroy habitats when the PESs were structured without credible deterrence of an early exit. Fortunately, the results of the experiment also suggest research to overcome these barriers by ensuring that PES financial incentives are scaled and structured to effectively promote conservation. The lessons from this study—though they issue from the particular context of this experiment—provide suggestions about how to structure benefit sharing schemes that could be used to promote conservation in a range of settings.

BlocTrain: Block-Wise Conditional Training and Inference for Efficient Spike-Based Deep Learning

Spiking neural networks (SNNs), with their inherent capability to learn sparse spike-based input representations over time, offer a promising solution for enabling the next generation of intelligent autonomous systems. Nevertheless, end-to-end training of deep SNNs is both compute- and memory-intensive because of the need to backpropagate error gradients through time. We propose BlocTrain, which is a scalable and complexity-aware incremental algorithm for memory-efficient training of deep SNNs. We divide a deep SNN into blocks, where each block consists of few convolutional layers followed by a classifier. We train the blocks sequentially using local errors from the classifier. Once a given block is trained, our algorithm dynamically figures out easy vs. hard classes using the class-wise accuracy, and trains the deeper block only on the hard class inputs. In addition, we also incorporate a hard class detector (HCD) per block that is used during inference to exit early for the easy class inputs and activate the deeper blocks only for the hard class inputs. We trained ResNet-9 SNN divided into three blocks, using BlocTrain, on CIFAR-10 and obtained 86.4% accuracy, which is achieved with up to 2.95× lower memory requirement during the course of training, and 1.89× compute efficiency per inference (due to early exit strategy) with 1.45× memory overhead (primarily due to classifier weights) compared to end-to-end network. We also trained ResNet-11, divided into four blocks, on CIFAR-100 and obtained 58.21% accuracy, which is one of the first reported accuracy for SNN trained entirely with spike-based backpropagation on CIFAR-100.

Towards Edge Computing Using Early-Exit Convolutional Neural Networks

In computer vision applications, mobile devices can transfer the inference of Convolutional Neural Networks (CNNs) to the cloud due to their computational restrictions. Nevertheless, besides introducing more network load concerning the cloud, this approach can make unfeasible applications that require low latency. A possible solution is to use CNNs with early exits at the network edge. These CNNs can pre-classify part of the samples in the intermediate layers based on a confidence criterion. Hence, the device sends to the cloud only samples that have not been satisfactorily classified. This work evaluates the performance of these CNNs at the computational edge, considering an object detection application. For this, we employ a MobiletNetV2 with early exits. The experiments show that the early classification can reduce the data load and the inference time without imposing losses to the application performance.

ON THE STATE OF SMALL AND MEDIUM-SIZED BUSINESSES IN RUSSIA

Particular attention is paid to the small and medium-sized enterprise (SME) sector, since it is this sector that is the driving force behind economic growth in the country, which is confirmed by a number of studies. The numerical superiority of small and medium-sized companies over large ones is obvious. Consequently, SMEs provide a significant share of the population with income and jobs. The quantitative growth of the sector contributes to the creation of new jobs, an increase in competition and labor productivity, an improvement in product quality, the development of innovations and an increase in the welfare of the population, that is, qualitatively favorable changes in the market and in the country's economy. Analysis of the state of the Russian SME sector, determination of its effectiveness and comparison with international practice is of great importance for drawing up a picture of differences, forming advantages and disadvantages, searching for options for its improvement. The relevance of the article lies in the fact that the development of the small and medium-sized business sector is one of the priority areas of state policy in Russia, therefore, stabilization of the situation and an early exit from the crisis period are very important for the Russian economy. The article discusses measures of state support for business that have been implemented in Russia since the beginning of the pandemic.

Economic, organizational and psychological determinants of early turnover: evidence from a pharmaceutical company in China

PurposeEarly turnover is a worldwide problem that occurs frequently during the first three years of employment. From a multidisciplinary perspective, this study attempts to find the economic, organizational and psychological factors that account for turnover at the early stage of employment.Design/methodology/approachThe authors used turnover records provided by the human resources division of a US pharmaceutical company operating in China of 222 Medical Representatives (MR). The method of Firth's logistic regression for analyzing was employed.FindingsAs an economic factor, the favorable labor conditions (i.e. high ratio of job vacancies) at the time of recruitment were inversely associated with MR subsequent retention. For organizational factors, unsatisfactory supervision and disappointment of intra-organizational career were the major predictors, and job ranks showed a U-shaped relationship to early resignation. Moreover, working pressure was a psychological factor of early exit.Originality/valueThis study provides organizations with empirical implications to devise retention plans for newcomers at risk of attrition, which prevent them from early turnover in the industry facing a talent shortage. Studies based on the company exit records have little been done in turnover literature.

Key Factors Influencing Talent Retention and Turnover in Convenience Stores: A Comparison of Managers’ and Employees’ Perspectives

This study identifies the main factors influencing turnover among convenience store employees from a managerial perspective and infers the changes necessary to reduce such high turnover rates. Employing the decision-making trial and evaluation laboratory (DEMATEL) methodology, it investigates the degree of mutual influence between evaluation indicators and constructs a network relation map for evaluation dimensions and criteria. This study also uses the DEMATEL-based analytic network process method to compute the influential weights of each dimension and criteria. According to the empirical results of the causality model, convenience store employees must first improve the relatedness dimension, and managers must first improve the existence dimension. These findings can help convenience store managers address talent retention and turnover problems, develop effective strategies to lower the high turnover rates at convenience stores, and offer solutions to new industry entrants to avoid potential problems that might lead to early exit.

Work-retirement transitions and mental health: A longitudinal analysis of the role of social protection generosity in 11 countries

Aims: This study aimed to analyse the effect of work-retirement transitions on post-retirement mental health in individuals with different working conditions in late working life. The focus was on transitions that involve the use of social protection schemes to bridge the gap between the exit from work and retirement, and the extent to which the generosity of such schemes is related to mental health after retirement. Methods: Individual-level panel data from the Survey of Health, Ageing and Retirement in Europe for 11 European countries were analysed using structural equation models. A total of 1642 individuals who worked in 2004 or 2007 and who retired in 2013 or 2015 were included in the analyses. The outcome measure was mental health as measured by the EURO-D scale. Results: Respondents with a ‘high strain’ and ‘passive’ work situation have a significantly higher likelihood of using social protection schemes, such as early retirement, sickness, disability and invalidity schemes before retirement. The generosity of such schemes has a significant positive relation to post-retirement mental health. Conclusions: This study shows that the generosity of early exit pathways is important for post-retirement mental health, especially for individuals with adverse working conditions at the end of their working lives.