Dave Snowden的动态

Save time analyzing MEA datasets with autoMEA. This Python package automates burst detection by using a machine learning model to predict optimal bursts parameters. It’s open source, inviting the MEA community to contribute and help improving it.

It is exciting to share our latest work which just got published in Science Robotics. This has been a collaborative effort between researchers from Universidad de Granada and EPFL. A big thank you to Alice Bruel, G. Courtine, Auke J. Ijspeert, Eduardo Ros, and Niceto Luque Sola, it is a pleasure working with you! In this work we inspire from neuromechanics to implement adjustable robot motor behavior. A cerebellar computational model provides motor adaptation and learning, thus obviating the need for any prior analytical modeling of the robot dynamics. It is combined with a muscle computational model that incorporates muscle viscoelasticity, variable cocontraction, and a reflex response based on spinal cord circuits. Modifying the cocontraction level regulates the robot stiffness, allowing to adjust the robot behavior depending on the environment and task requirements. High cocontraction enhances performance accuracy and increases robustness against external perturbations, while low cocontraction facilitates compliant human-robot physical interactions.

First paper at the IEEE/RSJ International Conference on Intelligent Robots and Systems: "Exploring 3D Human Pose Estimation and Forecasting from the Robot's Perspective: The HARPER Dataset". And first time that a dataset is collected using the onboard sensors of SPOT, the quadruped robot manufactured by Boston Dynamics (the model that remembers a mechanical dog). The key-challenge and novelty of the data is that the sensors are close to the ground and can capture the image of human users only partially, a new setting in Human-Robot Interaction. The work was possible thanks to the collaboration between the University of Verona (Italy) and the University of Glasgow (UK). Three students from Verona (Andrea Avogaro Federico Cunico and Andrea Toaiari), all supervised by @Marco Cristani , spent three months in Glasgow to collaborate with Xiangmin Xu and Haralambos Dafas (student of the Social AI CDT), both supervised by Emma Liying Li. A group of talented students and colleagues capable to overcome all technical obstacles, a privilege to work with them.

Nanobots inspired by the biology of bumblebees represent an exciting intersection between biology and nanotechnology. Bumblebees are remarkable for their complex sensory systems, efficient energy use, and sophisticated navigation skills, which have inspired scientists to explore similar abilities in nano-sized robots. Here’s how science draws on these traits in bumblebees to design nanobots: 1. Efficient Energy Use and Movement Bumblebees use energy efficiently, allowing them to hover and carry heavy pollen loads relative to their body size. Scientists study these biomechanics to create nanobots with optimized energy use, allowing them to perform tasks in tight or power-limited environments, like medical nanobots operating within the body or microscopic robots functioning in space. 2. Swarm Behavior Bumblebees exhibit complex swarm behavior, coordinating as a group to forage and protect their hive. Nanobot developers aim to mimic these swarm dynamics, enabling groups of tiny robots to work together to accomplish larger tasks, such as clearing clogged arteries, removing pollutants, or delivering drugs to targeted cells. 3. Sensory Adaptation Bumblebees navigate and detect flowers using complex sensory systems that analyze light, motion, and even electromagnetic fields. Nanobot researchers are inspired by this ability to develop sensory mechanisms that allow nanobots to detect specific environments or molecules. For instance, nanobots designed for medical purposes could locate cancer cells by identifying unique biomarkers. 4. Self-Healing Materials Bumblebees’ wings have a unique resilience, allowing minor tears to heal or remain functional without requiring immediate repair. Scientists are working on self-healing materials for nanobots to ensure that minor damages don’t hinder their functions, particularly in medical or environmental applications where reliable operation is critical. 5. Environmental Adaptation and Navigation Bumblebees adapt to various environments by learning and memorizing routes to flowers. In nanobot development, this translates to algorithms that allow nanobots to autonomously navigate complex environments. For example, a medical nanobot might navigate through the bloodstream to deliver medication precisely at an infection site. By studying bumblebees, researchers aim to create more sophisticated and resilient nanobots capable of complex, efficient, and adaptive functions that could one day transform fields like medicine, environmental cleanup, and technology. #Nanobots #BumblebeeScience #BioInspiredTech #Nanotechnology #SwarmIntelligence #MedicalNanobots #EnergyEfficiency #SelfHealingMaterials #Biomimicry #FutureOfTech #EnvironmentalAdaptation #ScienceInnovation #TinyTech

??It was a?true pleasure to share the stage with Junaid Bajwa, Sophie Ghazal, Teodora Popa, PhD, and Grant W Mitchell, MD, MBA at the ATOM Conference yesterday! Our panel explored the way AI is being adopted into the healthcare system, the barriers and enablers to this adoption, and the potential for these game-changing technologies. It’s not just about technology. With half the world still lacking access to essential healthcare, our discussion turned to equity. How can we ensure that algorithms are trained to deliver equitable care, when the datasets upon which they are trained are inherently biased? One particularly engaging topic was the integration of cutting-edge tools into operating rooms and clinics. Technology must meet clinicians where they are, enhancing workflows without adding complexity. Interoperability and clinician-first design are essential in ensuring adoption and effectiveness. A special thank you to Ted Souder for his excellent moderation and a huge thank you to Vartan Sarkissian for including me and Caresyntax in this incredible day at a truly special venue - Science Museum Group. Bjoern von Siemens David Cubbin Etienne Bauer Amel Mokrani Bois Pierre Mirlesse Gargee Kashyap Gulshan Batool

?? It’s always inspiring to see healthcare leaders tackling the big questions, and Bracy Fertig PhD’s panel at the ATOM Conference did exactly that. Personalising care in an increasingly digital world is not just a goal, but it’s becoming a necessity. The idea of AI reducing diagnostic errors by double digits and enabling faster, more accurate decision-making resonates deeply. But as Bracy highlights, the challenge goes beyond building the technology. We must ensure that it reaches every corner of the globe and is accessible to the general population. Equity in access to healthcare remains one of the most pressing issues of our time. The discussion on integrating AI into clinical was particularly insightful. Enhancing workflows without adding complexity, prioritising interoperability, and ensuring clinician-first design are crucial steps to adoption. After all, technology is only as good as the real-world problems it solves. A big thank you to the ATOM Conference for exploring big ideas and for reminding us all that meaningful progress requires collaboration, innovation, and action. ?? #DigitalHealth #PersonalizedCare #HealthcareInnovation #AIinHealthcare Dennis Kogan Etienne Bauer Vik Anantha Dan Juberg Michael Schelper Gulshan Batool Amel Mokrani Bois Matt Krueger David Cubbin Bracy Fertig PhD Gargee Kashyap Nicole Neal Taiki Esheim Bruce Ramshaw Caresyntax

Dear Network, I am thrilled to announce that my first research paper has been accepted at the 6th International Conference on Pattern Analysis and Intelligent Systems (PAIS), presented in April 2024, and added to IEEE Xplore on June 3, 2024. Abstract: This study investigates the use of a surrogate model based on deep neural networks to predict and estimate the travel time and energy consumption of a collaborative robot. Through the utilization of deep neural networks, two different Neural Network architecture have been tested and we conducted extensive simulations and evaluations. Our results indicate that the surrogate model’s predictions significantly outpace the computational time required by the original simulation itself. Furthermore, the accuracy of the predictions obtained from the surrogate model proved to be highly promising for the two models and particularly for the second proposal. This research underscores the efficacy of employing surrogate models in robotics applications, particularly in achieving faster and reliable estimations of system behavior. Such advancements hold significant implications for enhancing efficiency and decision making in robotics operations. Thank you to all the co-authors for their invaluable contributions: Mehdi Gaham, Bouchra KHOUMERI, Taqwa MEZAACHE, Afaf AMMAR, Isra RACHED.

Exciting Breakthroughs: Introducing AutoFilter and ClassSCT I am thrilled to announce the development of two groundbreaking platforms powered by machine learning: @AutoFilter and @ClassSCT. AutoFilter: Designed to revolutionize drug discovery, this platform processes a curated compound library of over 8 million molecules. It performs advanced shape- and chemistry-based complementarity analyses between ligands and binding pockets, delivering optimal drug-like candidates with exceptional precision. ClassSCT: A cutting-edge tool leveraging machine learning to achieve highly accurate estimations for detecting skin cancer from medical images. These innovations have been made possible thanks to the computational facilities of the University of Missouri and the Molecular Interactions Core at Mizzou. Additionally, I am deeply grateful to the talented students who contributed to the development of these platforms—congratulations to you all on this remarkable achievement! Stay tuned for more details about these transformative platforms and their potential to reshape research and healthcare!

2024?has?been?a?year?full?of?exciting?scientific?discoveries?and?breakthroughs!?Here?are?some?highlights: AI?in?Research?&?Development:?AI?continues?to?revolutionize?fields?like? chemistry?and?drug?discovery. Green?Chemistry:?Advances?in?greener?chemical?processes?are?helping?to? reduce?environmental?impact. Biomaterials:?New?biomaterials?are?being?developed?for?medical?and? industrial?applications. Exoplanet?Discovery:?NASA's?James?Webb?Space?Telescope?confirmed?its?first exoplanet,?LHS?475?b,?which?is?almost?the?same?size?as?Earth. CRISPR?Clinical?Validation:?CRISPR?technology?is?making?strides?in?clinical? applications,?particularly?in?treating?genetic?disorders. Decarbonizing?Energy:?Innovations?in?energy?production?are?focusing?on? reducing?carbon?emissions. Neurodegenerative?Disease?Treatments:?Progress?is?being?made?in?treating previously?"undruggable"?diseases?like?cancer?and?neurodegenerative? disorders. Unknown?Viruses:?New?viruses?have?been?discovered?in?everyday?items?like? toothbrushes?and?showerheads. It's?an?exciting?time?for?science!