nicola mariella’s Post

I will be at ICML in Vienna this week and presenting our paper on Quantum Contextual Optimal Transport (see below)! Hope to chat with many of you about AI4Science, Language Models & Quantum ML

✔️Quantum Learning and Shadows. ✔️Machine Learning for Quantum Science. ✔️Experimental Implementations. ✔️Architecture for #Quantum Machine Learning and more! Tune in to another exciting day at #QTML2023, hosted at CERN! 🔗https://lnkd.in/eyFTJF_y #CERNqti #quantumphysics #machinelearning

Just published in Nature Portfolio's Nature Communications: https://rdcu.be/dtHQg! 🔥🔥🔥 Our synergistic proposal shifts from the #quantum versus classical race and leverages the full power of tensor networks to use #QuantumComputing resources meaningfully. It also helps mitigate outstanding QPU trainability issues. We feel this paradigm is needed to show any practical quantum advantage in the future, independent of how long it will take us to reach this milestone and what type of quantum computer is used (e.g., NISQ vs. fault-tolerant, etc). Huge thanks to Manuel Rudolph, Jacob E. Miller, Danial Motlagh, Jing Chen, and Atithi Acharya for all their hard work, and to Zapata AI for funding this research. More details can be found in our paper, https://rdcu.be/dtHQg, or in the post below.

Thrilled to share that two of our latest research papers have been published in the conference proceedings of BioCAS 2023! "Feed-forward and recurrent inhibition for compressing and classifying high dynamic range biosignals in spiking neural network architectures" : A biologically-inspired approach that utilizes spike-frequency adaptation, feed-forward inhibition, and E-I balance effectively compresses high-dynamic range biosignals into spikes for SNN processing, demonstrating its feasibility in gesture classification tasks. Rachel Sava Giacomo Indiveri https://lnkd.in/dtfxR33G Congratulation to Rachel for the Charles A. Desoer Prize! "Online Unsupervised Arm Posture Adaptation for sEMG-based Gesture Recognition on a Parallel Ultra-Low-Power Microcontroller": An unsupervised adaptation technique for sEMG classification is proposed to improve generalization in arm posture variability scenarios. This approach achieves significant accuracy gains and can operate in real-time on embedded devices. Marcello Zanghieri Mattia Orlandi Emanuele Gruppioni Luca Benini and Simone Benatti. https://lnkd.in/dkdPaFcz

Quantum machine learning is a hot topic. In this work, collaborators, including members of QuEra’s team, explore the role that analog computing features—common in neutral-atom platforms—can play in these types of applications. Particular emphasis is given to the synergistic outcomes of analog and single-qubit gate-based digital capabilities operating in tandem. The findings in this work indicate that this hybrid operation mode can lead to shorter circuit depths and increased robustness to error. https://lnkd.in/gV3cV6Zn

I am thrilled to share that my previous work with Yen Jui Chang and other fellow quantum machine learning researchers has been accepted as a poster session at the 2024 IEEE #quantum week conference #quantumai #qubo #isingmachine #rbm #quantumml

Can variational #quantum #algorithms that exhibit barren plateaus still be trainable? Our work on synergistic pretraining of quantum circuits via tensor networks provides strong numerical evidence that this is possible, and it was just published in #Nature #Communications. Link to our paper: https://lnkd.in/eV5vHrwK In our work, we provide numerical evidence that deep quantum circuits that globally exhibit barren plateaus can scalably be pretrained to retain large gradients. Originally, we demonstrated this up to 20 qubits, but in the revision stage, we were able to push to 100 qubits. What was most fascinating for me was the insight that it may not require exponential classical resources to place a parametrized quantum circuit into a promising region in the loss landscape. This publication comes just in time for our collaboration’s recent paper, where we connected the provable absence of barren plateaus with efficient classical simulability. That work was led by Marco Cerezo and my Professor Zoë Holmes (here is the link: https://lnkd.in/ea-wa9Ga), and in combination with our now-published work highlights that we need to rethink the typical “classical vs quantum” dichotomy. It is increasingly clear that classical resources will play a crucial role in any future quantum advantage, and improvements on either side enhance the other. We thank the reviewers for their comments and for challenging us to continue providing convincing evidence. Big thanks to my former colleagues and friends Alejandro Perdomo-Ortiz, Jacob E. Miller, Jing Chen, Atithi Acharya, and special kudos to Danial Motlagh, who jumped onto the project during the revision stage and helped us push the envelope with the new Fig. 4.

We encourage you to read the recently published article, "Towards Scalable Digital Modeling of Networks of Biorealistic Silicon Neurons." 📖 It has results in OPEN SOURCE! 📝 Authored by: Swagat Bhattacharyya; Praveen Raj Ayyappan; Jennifer O. Hasler Volume: 13, Issue: 4, December 2023 Digital implementations of biorealistic neuron circuits for network computation have a trade-off between computational efficiency and biorealism. This work introduces efficient digital approximations for coupled Hodgkin-Huxley (HH) neurons using transistor-channel neural modeling and implements these models in C with both floating-point and 32-bit fixed-point arithmetic. This approach, which has been made open-source (https://loom.ly/7u6iNY4), allows for large-scale simulation of HH-like neurons, offering a scalable solution for digital modeling and paving the way for analog computing. 🔗 Read more on IEEE Xplore: https://loom.ly/7s34jWs 📖 This article has OPEN SOURCE results! https://loom.ly/7u6iNY4 #IEEE #IEEEXplore #JETCAS #PopularArticles #ReadMore #CircuitsandSystems #graphicalabstract #ReadMore #OpenSource

🚀Achieved Space Combatant in IBM Quantum Explorers! 🌌 ✨ Encoded data into a quantum computer 💡 Mastered supervised quantum machine learning (QSVM) 🔍 Explored quantum kernels, feature maps, and parameterized circuits 💻 Used Qiskit's machine learning library for hands-on experience The quantum future is now! #ibmquantum #QuantumAchievement #qiskit #SpaceCombatant 🚀#machinelearning #exploretheunknown

A study by the Barcelona Institute of Science and Technology (BIST) and partners used attention-based neural networks to improve quantum state tomography by denoising experimental data. Quantum Brilliance and Oak Ridge National Laboratory have partnered to integrate room-temperature diamond quantum accelerators with high-performance computing systems. A study by CSIRO and the University of Melbourne demonstrates that quantum principal component analysis effectively compresses data from chemiresistive sensor arrays. And much more. Read every qubit 👉 https://lnkd.in/gfHXp--K #QuantumComputing #QuantumAlgorithms #QuantumInformation #Innovation