Audio applications involving environmental sound analysis increasingly use general-purpose audio representations, also known as embeddings, for transfer learning. Recently, Holistic Evaluation of Audio Representations (HEAR) evaluated twenty-nine embedding models on nineteen diverse tasks. However, the evaluation's effectiveness depends on the variation already captured within a given dataset. Therefore, for a given data domain, it is unclear how the representations would be affected by the variations caused by myriad microphones' range and acoustic conditions -- commonly known as channel effects. We aim to extend HEAR to evaluate invariance to channel effects in this work. To accomplish this, we imitate channel effects by injecting perturbations to the audio signal and measure the shift in the new (perturbed) embeddings with three distance measures, making the evaluation domain-dependent but not task-dependent. Combined with the downstream performance, it helps us make a more informed prediction of how robust the embeddings are to the channel effects. We evaluate two embeddings -- YAMNet, and OpenL3 on monophonic (UrbanSound8K) and polyphonic (SONYC-UST) urban datasets. We show that one distance measure does not suffice in such task-independent evaluation. Although Fr\'echet Audio Distance (FAD) correlates with the trend of the performance drop in the downstream task most accurately, we show that we need to study FAD in conjunction with the other distances to get a clear understanding of the overall effect of the perturbation. In terms of the embedding performance, we find OpenL3 to be more robust than YAMNet, which aligns with the HEAR evaluation.

This article presents new hexagonal and pentagonal PEM fuel cell models. The models have been optimized after achieving improved cell performance. The input parameters of the multi-objective optimization algorithm were pressure and temperature at the inlet, and consumption and output powers were the objective parameters. The output data of the numerical simulation has been trained using deep neural networks and then modeled with polynomial regression. The target functions have been extracted using the RSM (Response Surface Method), and the targets were optimized using the multi-objective genetic algorithm (NSGA-II). Compared to the base model, the optimized Pentagonal and Hexagonal models increase the output current density by 21.8% and 39.9%, respectively.

Spiking neural network (SNN) has been attached to great importance due to the properties of high biological plausibility and low energy consumption on neuromorphic hardware. As an efficient method to obtain deep SNN, the conversion method has exhibited high performance on various large-scale datasets. However, it typically suffers from severe performance degradation and high time delays. In particular, most of the previous work focuses on simple classification tasks while ignoring the precise approximation to ANN output. In this paper, we first theoretically analyze the conversion errors and derive the harmful effects of time-varying extremes on synaptic currents. We propose the Spike Calibration (SpiCalib) to eliminate the damage of discrete spikes to the output distribution and modify the LIPooling to allow conversion of the arbitrary MaxPooling layer losslessly. Moreover, Bayesian optimization for optimal normalization parameters is proposed to avoid empirical settings. The experimental results demonstrate the state-of-the-art performance on classification, object detection, and segmentation tasks. To the best of our knowledge, this is the first time to obtain SNN comparable to ANN on these tasks simultaneously. Moreover, we only need 1/50 inference time of the previous work on the detection task and can achieve the same performance under 0.492$\times$ energy consumption of ANN on the segmentation task.

Argonne National Laboratory, a multidisciplinary science and engineering research centre operated by the University of Chicago for the U.S. Department of Energy, has installed Graphcore's Intelligence Processing Units (IPUs) within its AI Testbed. The AI Testbed, operated by the Argonne Leadership Computing Facility (ALCF), enables researchers to explore next-generation machine learning applications and workloads to advance the use of AI for science. Technologies selected for the testbed, such as Graphcore's IPUs, complement the facility's current and next-generation supercomputers to provide a state-of-the-art environment that supports pioneering research at the intersection of AI, big data, and high-performance computing (HPC). The systems in the ALCF AI Testbed are purpose-built for machine learning and data-centric workloads, making them well suited to address challenges involving the increasingly large amounts of data produced by supercomputers, light sources, and particle accelerators among other powerful research tools. In addition, the testbed allows researchers to explore novel workflows that combine AI methods with simulation and experimental science to accelerate the pace of discovery.

In March, a troop of engineers gathered in an unkept green field in rural Nottinghamshire, England. They were there to test a drone piloting software that they hoped could one day be in charge of maintaining the high-voltage pylons that transmit electricity across the country. Assuming the software was working, a drone was about to inspect a pylon from a few meters away, maneuvered not by a nearby pilot but a computer in a control station hundreds of meters away. Seconds later, the dance began. Whizzing around, the drone took 65 photos that documented the condition of the pylon's steel arms, fittings, and conductors.

LG Energy Solution announced the launch of its Advisory Council on Artificial Intelligence (AI). Joining forces with leading scholars specializing in AI technology, the Advisory Council is expected to play a major role in LGES's digital transformation and the establishment of a manufacturing intelligence platform. The newly formed council will set visions and business directions to enhance the company's digital transformation, as well as build the technological partnerships required to realize these ambitions. The company has appointed five AI experts as its committee members: Sungroh Yoon (Ph.D, Electrical Engineering, Stanford University), Byung-Gon Chun (Ph.D, Computer Science, University of California, Berkeley), Jinwoo Shin (Ph.D, Mathematics, Massachusetts Institute of Technology), Frank Chongwoo Park (Ph.D, Applied Mathematics, Harvard University), and Jong Min Lee (Ph.D, Chemical Engineering, Georgia Institute of Technology). Each advisory member will be assigned to a different division, and oversee the selection and execution of strategic assignments within their respective area of expertise.

Echo state network (ESN), a kind of recurrent neural networks, consists of a fixed reservoir in which neurons are connected randomly and recursively and obtains the desired output only by training output connection weights. First-order reduced and controlled error (FORCE) learning is an online supervised training approach that can change the chaotic activity of ESNs into specified activity patterns. This paper proposes a composite FORCE learning method based on recursive least squares to train ESNs whose initial activity is spontaneously chaotic, where a composite learning technique featured by dynamic regressor extension and memory data exploitation is applied to enhance parameter convergence. The proposed method is applied to a benchmark problem about predicting chaotic time series generated by the Mackey-Glass system, and numerical results have shown that it significantly improves learning and prediction performances compared with existing methods.

Experts in the United Kingdom are designing AI-powered robots to work in some of the most hazardous places on Earth and outer space. University of Manchester researchers have been advising government and energy sector leaders on the safe development of AI robots being used in extreme environments. The University of Manchester said "hot robotic" systems were originally designed to work in radioactive environments in decommissioned nuclear reactors, but found a new use in the fields of nuclear fusion power, agriculture, the energy sector, and even space exploration. The university said in a statement: "As part of an ambitious R&D program to maintain UK leadership in robotic technologies, Manchester experts are applying AI technologies to'hot robotics' as they will increasingly need to act independently of human operators as they enter a range of danger zones to carry out highly complex tasks. "An important challenge in the nuclear industry is to improve robot autonomy so that the technology can be used to deliver safer, faster and cheaper decommissioning of legacy power stations and other radioactive facilities at sites such as Sellafield and Dounreay.

New Delhi: Centre of Excellence on Satellite & Unmanned Remote Vehicle Initiative (CoE-SURVEI) has developed an Artificial Intelligence-based software which can automatically detect change on ground, including unauthorised constructions and encroachments in a time series using Satellite Imagery. The CoE-SURVEI, established by Directorate General Defence Estates (DGDE) at National Institute of Defence Estates Management, leverages latest technologies in survey viz. The CoE was inaugurated by Raksha Mantri Rajnath Singh on December 16, 2021. This Change Detection Software has been developed by CoE-SURVEI in collaboration with knowledge partner Bhabha Atomic Research Centre (BARC), Visakhapatnam. Presently, the tool uses National Remote Sensing Centre (NRSC) Cartosat-3 imagery with trained software.

The objective of model selection is to find the network architecture with the best generalization properties. We want to improve the final selection error obtained before (0.263 NSE). The best selection error is achieved using a model with the most appropriate complexity to produce a good data fit. Order selection algorithms are responsible for find the optimal number of perceptrons in the neural network. The following chart shows the results of the incremental order algorithm.