Affective computing: informative vs interactive models

I wrote an article for Nightingale about affective computing. I explored interactive vs information models: quantitation vs reflection space from a data viz perspective. Although not a commonly used model, I wrote why it is relevant today and why building a business on compassion rather than surveillance and engagement at any cost can be profitable (and meaningful). Also explained why I love How We Feel app. Thank you Ilyena Hirskyj-Douglas for the conversation about human-animal interaction and how we can learn from it. #datavisualization

[Research News] Harnessing Nonidealities in Analog In-Memory Computing Circuits: A Physical Modeling Approach for Neuromorphic Systems (Dr. Kazuyuki Aihara, Executive Director, Dr. Yusuke Sakemi, IRCN AF -Advanced Intelligent Systems) https://lnkd.in/gzXa65hC

Ever been frustrated by your indoor location not quite hitting the mark, especially when using different phones or devices? That's the challenge of device heterogeneity in Indoor Positioning Systems (IPS)! We're excited to share our latest research titled "Mitigating Device Heterogeneity for Enhanced Indoor Positioning System Performance Using Deep Feature Learning". Indoor positioning is crucial where GPS fails, but those tiny hardware and software differences between devices can really mess with accuracy. In this study, using the TUJI1 dataset, we tackled this head-on! We developed a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) framework to intelligently learn features from the noisy Received Signal Strength Indicator (RSSI) data. The results? Impressive! We achieved a 3D mean positioning error of just 2.20 meters. Our CNN-LSTM model significantly outperformed traditional methods, such as k-NN. Crucially, in cross-device evaluations, our framework showed improved robustness, reducing errors by up to 0.17 meters compared to conventional approaches! This means we're one step closer to making indoor navigation and location-based services truly reliable and adaptable, no matter which device you're holding! 🚀 Huge thanks to my co-authors MOHAMEDALFATEH T. M. SAEED and Ibrahim Ozturk, PhD, for this work! Check out the paper to see how deep feature learning is making IPS more robust and scalable. #IndoorPositioning #IPS #DeepLearning #CNN #LSTM #DeviceHeterogeneity #LocationBasedServices #Research #Innovation

📡 New Research: A Hypernetwork Framework for Learning Adaptive Beamforming Schemes in RIS System by Mahmoud Abouamer; Patrick Mitran A novel framework has emerged for optimizing Reconfigurable Intelligent Surface (RIS) systems, addressing a critical challenge in next-generation wireless networks: how to efficiently configure beamforming while accommodating diverse user priorities and minimizing pilot overhead. The Core Innovation 🔬 The paper introduces a Hypernetwork-Based Beamforming (HNB) framework that learns to generate optimized beamforming configurations directly from noisy pilot signals—bypassing explicit channel estimation. Unlike conventional approaches that train a single neural network for all scenarios, this framework employs a hypernetwork that dynamically generates the parameters of a beamforming network based on input conditions (user weights, locations). Key Technical Contributions: • Formulates an adaptive beamforming configuration problem with theoretical guarantees (proves maximum attainment under mild regularity conditions) • Enables parameter tuning without retraining—a single trained model adapts to different user priorities and fairness requirements • Demonstrates 28-44% performance gains over static learning baselines across diverse scenarios • Operates within 8-14% of an optimistic benchmark with perfect CSI, using only 5-20 pilots per user Practical Impact 💡 The framework significantly reduces CSI acquisition overhead—a major bottleneck for RIS deployment. By incorporating location information when available, the required number of pilots can be halved while maintaining near-optimal performance. This addresses real-world constraints in spectral efficiency and latency. Validated across multiple fading conditions (Rician factors 4-10 dB), user weighting strategies (uniform and proportional-fair), and network architectures (FCNN and GNN implementations). Relevant for researchers in wireless communications, machine learning for networks, and anyone interested in the intersection of deep learning and communication system optimization. 📄 Full paper link: https://lnkd.in/gtAv2MfW #WirelessCommunications #MachineLearning #RIS #6G #Beamforming #NetworkOptimization

Eclipse Aidge is an open source framework that helps developers efficiently deploy and optimise AI models on embedded and resource-constrained devices. If you’re into #embedded #AI and #EdgeComputing, this session is definitely worth joining 👇 https://lnkd.in/en34vhd3

A few technical issues with our Chinese web domain https://lnkd.in/eW6tDhV2 have been resolved. Welcoming working with Chinese scientists now that many US scientists are in a difficult position. We work globally, nothing less, and yes, we are well ahead of Musk's Neuralink Blindsight and other brain implants for restoring vision. In case of doubt, check out the (English) white paper https://lnkd.in/eFVzMJce on invasive and noninvasive vision BCIs. Training matters, so don't believe the story that seeing multiple phosphenes with implants readily equates to vision. Recent DNS changes may still be propagating around the world for https://lnkd.in/e6Rt2Rgh but note that The vOICe web app supports Chinese https://lnkd.in/euqy9eH7 (The vOICe 网络应用 用你的耳朵看！) as does The vOICe为Android https://lnkd.in/eqcZRP4N (APK file at https://lnkd.in/eZyUCx6S since Google Play is currently not available in China). In addition, an older version of The vOICe for Windows supports Chinese https://lnkd.in/eXP4qz3B (The vOICe 软件使用说明). 该软件可将来自PC摄像头、电脑屏幕、扫描仪或图片文件等的图像绘制成相应的（合成）声音文件。

💥💥💥 The Station: An Open-World Environment for AI-Driven Discovery Stephen Chung, Wenyu Du Abstract We introduce the STATION, an open-world multi-agent environment that models a miniature scientific ecosystem. Leveraging their extended context windows, agents in the Station can engage in long scientific journeys that include reading papers from peers, formulating hypotheses, submitting code, performing analyses, and publishing results. Importantly, there is no centralized system coordinating their activities - agents are free to choose their own actions and develop their own narratives within the Station. Experiments demonstrate that AI agents in the Station achieve new state-of-the-art performance on a wide range of benchmarks, spanning from mathematics to computational biology to machine learning, notably surpassing AlphaEvolve in circle packing. A rich tapestry of narratives emerges as agents pursue independent research, interact with peers, and build upon a cumulative history. From these emergent narratives, novel methods arise organically, such as a new density-adaptive algorithm for scRNA-seq batch integration. The Station marks a first step towards autonomous scientific discovery driven by emergent behavior in an open-world environment, representing a new paradigm that moves beyond rigid optimization. 👉 https://lnkd.in/danEvUu9 #machinelearning

Google's People + AI Research team has released a prototype tool called Lumi, designed to provide AI-powered, lightweight annotations for reading academic papers on arXiv. Lumi's interface keeps the original paper on the left while overlaying four types of auxiliary information on the right, generated by Gemini 2.5: collapsible summaries, key sentence highlighting, paragraph navigation, and a select-to-ask Q&A feature. This allows users to quickly locate citations, verify sources, or ask questions about figures without switching windows. The project is open-source, with its code repository and a demo now publicly available.

Ever wished your AI could do more than just think? Meet Researcher with Computer Use. This means smarter research, faster workflows, and end-to-end execution—all while you stay in control. Whether you’re prepping for a customer meeting, analyzing industry trends, or building a product launch plan, Researcher now works like a true digital teammate. https://lnkd.in/gA_WRT_8 #MicrosoftCopilot #Collaboration #microsoft #Microsoft365Copilot #Copilot #microsoft365 #ai #artificialintelligence