Bio-inspired spiking neural networks (SNNs), operating with asynchronous binary signals (or spikes) distributed over time, can potentially lead to greater computational efficiency on event-driven hardware. The state-of-the-art SNNs suffer from high inference latency, resulting from inefficient input encoding, and sub-optimal settings of the neuron parameters (firing threshold, and membrane leak). We propose DIET-SNN, a low latency deep spiking network that is trained with gradient descent to optimize the membrane leak and the firing threshold along with other network parameters (weights). The membrane leak and threshold for each layer of the SNN are optimized with end-to-end backpropagation to achieve competitive accuracy at reduced latency. The analog pixel values of an image are directly applied to the input layer of DIET-SNN without the need to convert to spike-train. The information is converted into spikes in the first convolutional layer where leaky-integrate-and-fire (LIF) neurons integrate the weighted inputs and generate an output spike when the membrane potential crosses the trained firing threshold. The trained membrane leak controls the flow of input information and attenuates irrelevant inputs to increase the activation sparsity in the convolutional and linear layers of the network. The reduced latency combined with high activation sparsity provides large improvements in computational efficiency. We evaluate DIET-SNN on image classification tasks from CIFAR and ImageNet datasets on VGG and ResNet architectures. We achieve top-1 accuracy of 66.52% with 25 timesteps (inference latency) on the ImageNet dataset with 3.1X less compute energy than an equivalent standard ANN. Additionally, DIET-SNN performs 5-100X faster inference compared to other state-of-the-art SNN models.

Among the many departments and agencies within the United States federal government, the US Department of Energy (DOE) stands out as one of the most science, technology, and innovation-focused. This should come as little surprise to those who know the DOE's storied history with its breakthrough labs, world-leading research institutions, and highly educated staff. Since World War II, the DOE has been at the forefront of most of the groundbreaking and world-changing revolutions in science and technology including the development and harnessing of nuclear energy, innovations in genomics including the DOE initiative Human Genome Project, work in high-performance computing, and many other research-oriented efforts. In fact, the DOE supports more research in the physical sciences than any other US federal agency, providing more than 40% of US funding in computing, physics, chemistry, materials science, and other area through a system of national laboratories including Lawrence Berkeley National Laboratory, Oak Ridge National Laboratory, Argonne National Laboratory, Ames Laboratory, Brookhaven National Laboratory, Los Alamos National Laboratory, Sandia National Labs, Lawrence Livermore National Laboratory, the SLAC National Accelerator Laboratory, and dozens more institutions. Until very recently, the DOE also ran the world's top two fastest supercomputers: Summit and Sierra.

According to reports, India has witnessed an exponential growth in interest and awareness in the solar energy space, and this has led to growth in the number of solar plant installations and a decrease in the cost of solar power. With the aim of using AI and cloud computing into solar energy platforms, Noida-based Skilancer Solar manufactures centrally controlled, self-powered, robotic arms for automatic cleaning of solar modules. In this week's feature, Analytics India Magazine caught up with the founders of Skilancer Solar to gain more insights into the clean-tech platform. Founded in 2018 by Manish Das and Neeraj Kumar, Skilancer Solar is a clean-tech startup that manufactures robots to clean solar panels. A lot of industries in India utilise traditional/manual methods of cleaning which are less optimal, use water inefficiently, require manpower and the cleaning frequency is also less, this often results in decreased power output by the solar plants.

Russian mercenary groups have enabled renegade military commander Khalifa Hafter, who is based in eastern Libya, to blockade the country's oil exports, starving the country of much-needed money. Moscow's backing of Haftar, a former CIA asset, has increased tensions with the United States. Russian private military contractors are active in 16 African nations. How is the country paying for its overseas wars? Also on Counting the Cost: Currency crisis, debt default, hyperinflation and poverty - Lebanon was in economic and political paralysis long before the devastating explosion in Beirut.

I'm not sure if Moflin is supposed to be a robotic hamster, guinea pig, baby bunny, or some alternate take on a Tribble, but goddamn this robo-pet is cute. Launched as part of a Kickstarter campaign from Vanguard Industries that went live earlier this week, Moflin looks to follow in the steps of Sony's Aibo or other robo-pets like Qoobo. However, instead of simply a disembodied tail attached to a fluffy base like Qoobo, Moflin apparently uses AI to have "emotional capabilities" meant to more accurately mimic real pets, so that it can express feelings and potentially even serve as a therapeutic aid. In order to make that happen, Vanguard Industries said it created its own Emotion AI tech that allows Moflin's feelings to react and evolve over time based on contact with humans. Individual Moflins can even have unique personalities based on their experiences, and learn to react differently depending on the actions of their owners.

This paper introduces structured machine learning regressions for prediction and nowcasting with panel data consisting of series sampled at different frequencies. Motivated by the empirical problem of predicting corporate earnings for a large cross-section of firms with macroeconomic, financial, and news time series sampled at different frequencies, we focus on the sparse-group LASSO regularization. This type of regularization can take advantage of the mixed frequency time series panel data structures and we find that it empirically outperforms the unstructured machine learning methods. We obtain oracle inequalities for the pooled and fixed effects sparse-group LASSO panel data estimators recognizing that financial and economic data exhibit heavier than Gaussian tails. To that end, we leverage on a novel Fuk-Nagaev concentration inequality for panel data consisting of heavy-tailed $\tau$-mixing processes which may be of independent interest in other high-dimensional panel data settings.

Much has been said about the fusion of bio-inspired optimization algorithms and Deep Learning models for several purposes: from the discovery of network topologies and hyper-parametric configurations with improved performance for a given task, to the optimization of the model's parameters as a replacement for gradient-based solvers. Indeed, the literature is rich in proposals showcasing the application of assorted nature-inspired approaches for these tasks. In this work we comprehensively review and critically examine contributions made so far based on three axes, each addressing a fundamental question in this research avenue: a) optimization and taxonomy (Why?), including a historical perspective, definitions of optimization problems in Deep Learning, and a taxonomy associated with an in-depth analysis of the literature, b) critical methodological analysis (How?), which together with two case studies, allows us to address learned lessons and recommendations for good practices following the analysis of the literature, and c) challenges and new directions of research (What can be done, and what for?). In summary, three axes - optimization and taxonomy, critical analysis, and challenges - which outline a complete vision of a merger of two technologies drawing up an exciting future for this area of fusion research.

In 2022, after a series of upgrades, CERN's Large Hadron Collider (LHC) is expected to turn back on for a final run and, once again, furiously smash particles together in search of new clues about the fundamental structure of our universe. When the LHC turns on, it will be operating at its highest energies yet: the collider will crash together protons every 25 nanoseconds, leading to the production of hundreds of particles passing through the detector. To help with the deluge of data, scientists are turning to new artificial intelligence (AI) technologies. We met with Jennifer Ngadiuba, a Robert A. Millikan Postdoctoral Scholar Research Associate in Physics at Caltech, over Zoom to learn more about these developments. Ngadiuba is also supported by the U.S. Department of Energy through Fermilab's Machine Intelligence group.

Underlying the current ciris, climate change remains a pressing concern. The current coronavirus pandemic has quickly demanded global attention as policymakers race to find a safe exit strategy. But as we pour our collective efforts into tackling this new threat, we must not forget about other invisible enemies that require our urgent attention. The climate emergency, for one, is a hugely complex global challenge with long-term implications. And we are pressed on time to come up with ways of reversing the damage done by humanity to the planet.

Distributed linguistic representations are powerful tools for modelling the uncertainty and complexity of preference information in linguistic decision making. To provide a comprehensive perspective on the development of distributed linguistic representations in decision making, we present the taxonomy of existing distributed linguistic representations. Then, we review the key elements of distributed linguistic information processing in decision making, including the distance measurement, aggregation methods, distributed linguistic preference relations, and distributed linguistic multiple attribute decision making models. Next, we provide a discussion on ongoing challenges and future research directions from the perspective of data science and explainable artificial intelligence.