No abstract

The recent popularity of deep neural networks (DNNs) has generated a lot of research interest in performing DNN-related computation efficiently. However, the primary focus is usually very narrow and limited to (i) inference -i.e. how to efficiently execute already trained models and (ii) image classification networks as the primary benchmark for evaluation.Our primary goal in this work is to break this myopic view by (i) proposing a new benchmark for DNN training, called TBD 1 , that uses a representative set of DNN models that cover a wide range of machine learning applications: image classification, machine translation, speech recognition, object detection, adversarial networks, reinforcement learning, and (ii) by performing an extensive performance analysis of training these different applications on three major deep learning frameworks (TensorFlow, MXNet, CNTK) across different hardware configurations (single-GPU, multi-GPU, and multi-machine). TBD currently covers six major application domains and eight different state-of-the-art models. We present a new toolchain for performance analysis for these models that combines the targeted usage of existing performance analysis tools, careful selection of new and existing metrics and methodologies to analyze the results, and utilization of domain specific characteristics of DNN training. We also build a new set of tools for memory profiling in all three major frameworks; much needed tools that can finally shed some light on precisely how much memory is consumed by different data structures (weights, activations, gradients, workspace) in DNN training. By using our tools and methodologies, we make several important observations and recommendations on where the future research and optimization of DNN training should be focused.

No abstract

Biomonitoring of effects in agricultural workers is necessary to assess the individual risk of handling pesticides. In this study, biochemical and haematological parameters were measured to evaluate the effects of exposure to these compounds in agricultural workers. The study was carried out in 110 workers and 97 control subjects. Several haematological and biochemical parameters were analysed. Assessment of haematological parameters revealed that the mean cell volume and haematocrit levels were significantly lower in workers than in controls (P ¼ 0.002 and 0.013, respectively), while mean corpuscular haemoglobin concentrations were higher in workers (Po0.001). There was also a significant inhibition of butyrylcholinesterase activity in workers compared with that in controls (Po0.001). Assessment of biochemical parameters further showed significantly higher activities of transferases, lactate dehydrogenase (Po0.001), alkaline phosphatase (ALP) (P ¼ 0.006) and creatine kinase (CK) (Po0.015), as well as higher levels of proteins (Po0.001), creatinine (P ¼ 0.001) and urea (P ¼ 0.001) in workers compared with controls, along with significantly higher uric acid levels (P ¼ 0.012). Furthermore, the number of years exposed to pesticides predicted higher activities of alanine aminotransferase, CK, ALP, as well as uric acid levels. Overall, chronic exposure to pesticides appeared to affect several biochemical parameters. These biomarkers seem to be indicative of adverse effects of pesticides in agricultural workers, confirming their use for routine monitoring of effects.

Epidemiological studies suggest that cytogenetic biomarkers, such as micronuclei (MN) in peripheral blood lymphocytes may predict cancer risk because they indicate genomic instability. The objective of the present study was to evaluate the frequencies of MN and chromosome aberrations (CA) in peripheral blood lymphocytes of hospital workers exposed to ionizing radiation and healthy subjects. The study was conducted using peripheral blood lymphocytes from 30 workers from the radiology department and 30 from the cardiology department. This study included 27 healthy age- and sex-matched individuals as the control group. The assessment of chromosomal damage was carried out by the use of CA and micronucleus assays in peripheral lymphocytes. Our results show that CA and micronucleus frequencies were significantly higher among the exposed groups when compared to controls. Our finding of significant increase of CA and MN frequencies in peripheral lymphocytes in exposed workers indicates a potential cytogenetic hazard due to this exposure. The enhanced chromosomal damage of subjects exposed to genotoxic agents emphasizes the need to develop safety programs.