A Novel Hybrid Cache Coherence with Global Snooping for Many-core Architectures

Cache coherence ensures correctness of cached data in multi-core processors. Traditional implementations of existing protocols make them unscalable for many core architectures. While snoopy coherence requires unscalable ordered networks, directory coherence is weighed down by high area and energy overheads. In this work, we propose Wireless-enabled Share-aware Hybrid (WiSH) to provide scalable coherence in many core processors. WiSH implements a novel Snoopy over Directory protocol using on-chip wireless links and hierarchical, clustered Network-on-Chip to achieve low-overhead and highly efficient coherence. A local directory protocol maintains coherence within a cluster of cores, while coherence among such clusters is achieved through global snoopy protocol. The ordered network for global snooping is provided through low-latency and low-energy broadcast wireless links. The overheads are further reduced through share-aware cache segmentation to eliminate coherence for private blocks. Evaluations show that WiSH reduces traffic by and runtime by , while requiring smaller storage and lower energy as compared to existing hierarchical and hybrid coherence protocols. Owing to its modularity, WiSH provides highly efficient and scalable coherence for many core processors.

Wind and wave regime over the Bombay high area under the influence of cyclonic storms

Temporal distributions of wind and wave over Bombay High Area (BHA) during cyclone period have been studied. Ten years’ (1990-99) data of BHA during cyclone period have been used. It is found that under the influence of cyclonic storms strong southwesterly winds prevail over the BHA in pre-monsoon and weaker east to southeasterly winds during post-monsoon. Southwesterly wave with heights exceeding 20 feet are encountered in BHA during pre-monsoon and south easterlies with wave height reaching up to 12 feet in post monsoon. Analysis of situations with different storm locations also yielded similar results. Relationships between wind speeds and wave height as well as the distance of the storm centre over BHA have been established.

Unstructured clinical notes within the 24 hours since admission predict short, mid & long-term mortality in adult ICU patients

Mortality prediction for intensive care unit (ICU) patients is crucial for improving outcomes and efficient utilization of resources. Accessibility of electronic health records (EHR) has enabled data-driven predictive modeling using machine learning. However, very few studies rely solely on unstructured clinical notes from the EHR for mortality prediction. In this work, we propose a framework to predict short, mid, and long-term mortality in adult ICU patients using unstructured clinical notes from the MIMIC III database, natural language processing (NLP), and machine learning (ML) models. Depending on the statistical description of the patients’ length of stay, we define the short-term as 48-hour and 4-day period, the mid-term as 7-day and 10-day period, and the long-term as 15-day and 30-day period after admission. We found that by only using clinical notes within the 24 hours of admission, our framework can achieve a high area under the receiver operating characteristics (AU-ROC) score for short, mid and long-term mortality prediction tasks. The test AU-ROC scores are 0.87, 0.83, 0.83, 0.82, 0.82, and 0.82 for 48-hour, 4-day, 7-day, 10-day, 15-day, and 30-day period mortality prediction, respectively. We also provide a comparative study among three types of feature extraction techniques from NLP: frequency-based technique, fixed embedding-based technique, and dynamic embedding-based technique. Lastly, we provide an interpretation of the NLP-based predictive models using feature-importance scores.

The individual and combined associations of depression and socioeconomic status with risk of major cardiovascular events: a prospective cohort study

Objective: We aimed to investigate the individual and combined associations of depression and low socioeconomic status (SES) with risk of major cardiovascular events (MCVE), defined as first-ever fatal or non-fatal stroke or myocardial infarction, in a large prospective cohort study. Methods: We used data from 466,238 UK Biobank participants, aged 40 - 69 years without cardiovascular disease, bipolar disorder or schizophrenia at baseline. We performed Cox proportional hazard models to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) of the individual and combined associations of depression and each of educational attainment, area-based deprivation and income with risk of MCVE. We assessed effect modification and explored interaction on the additive and multiplicative scale. Results: Depression, low education, high area-based deprivation and low income were individually associated with increased risks of MCVE (adjusted HR, 95% CI: 1.28, 1.19 - 1.38; 1.20, 1.14 - 1.27; 1.17, 1.11 - 1.23; and 1.22, 1.16 - 1.29, respectively). Depression was associated with increased risks of MCVE among individuals with high and low SES. Individuals with depression and each of low education, high area-based deprivation and low income were at particularly high risk of MCVE (HR, 95% CI: 1.50, 1.38 - 1.63; 1.63, 1.46 - 1.82; 1.31, 1.23 - 1.40, respectively). There was interaction between depression and area-based deprivation on multiplicative and additive scales but no interaction with education or income. Conclusion: Depression was associated with increased risks of MCVE among individuals with high and low SES, with particularly high risks among those living in areas of high deprivation.

Zero-Aware Low-Precision RNS Scaling Scheme

Scaling is one of the complex operations in the Residue Number System (RNS). This operation is necessary for RNS-based implementations of deep neural networks (DNNs) to prevent overflow. However, the state-of-the-art RNS scalers for special moduli sets consider the 2k modulo as the scaling factor, which results in a high-precision output with a high area and delay. Therefore, low-precision scaling based on multi-moduli scaling factors should be used to improve performance. However, low-precision scaling for numbers less than the scale factor results in zero output, which makes the subsequent operation result faulty. This paper first presents the formulation and hardware architecture of low-precision RNS scaling for four-moduli sets using new Chinese remainder theorem 2 (New CRT-II) based on a two-moduli scaling factor. Next, the low-precision scaler circuits are reused to achieve a high-precision scaler with the minimum overhead. Therefore, the proposed scaler can detect the zero output after low-precision scaling and then transform low-precision scaled residues to high precision to prevent zero output when the input number is not zero.

ModuloNET: Neural Networks Meet Modular Arithmetic for Efficient Hardware Masking

Intellectual Property (IP) thefts of trained machine learning (ML) models through side-channel attacks on inference engines are becoming a major threat. Indeed, several recent works have shown reverse engineering of the model internals using such attacks, but the research on building defenses is largely unexplored. There is a critical need to efficiently and securely transform those defenses from cryptography such as masking to ML frameworks. Existing works, however, revealed that a straightforward adaptation of such defenses either provides partial security or leads to high area overheads. To address those limitations, this work proposes a fundamentally new direction to construct neural networks that are inherently more compatible with masking. The key idea is to use modular arithmetic in neural networks and then efficiently realize masking, in either Boolean or arithmetic fashion, depending on the type of neural network layers. We demonstrate our approach on the edge-computing friendly binarized neural networks (BNN) and show how to modify the training and inference of such a network to work with modular arithmetic without sacrificing accuracy. We then design novel masking gadgets using Domain-Oriented Masking (DOM) to efficiently mask the unique operations of ML such as the activation function and the output layer classification, and we prove their security in the glitch-extended probing model. Finally, we implement fully masked neural networks on an FPGA, quantify that they can achieve a similar latency while reducing the FF and LUT costs over the state-of-the-art protected implementations by 34.2% and 42.6%, respectively, and demonstrate their first-order side-channel security with up to 1M traces.