Tudor Morosan, PhD’s Post

I’m happy to share that I’ve successfully completed the QBronze workshop and obtained a new certification: QBronze136 workshop organized by QWorld & WOMANIUM as part of Womanium Global Quantum + AI Program 2024 What I learnt: > Basics of Classical and Quantum Systems > Quantum Protocols and Algorithms > Qiskit introduction > Grover's Search Algorithm #Womanium #QWorld #QBronze

AI and quantum machines will converge at some point. When they do things are going to get really interesting.. https://lnkd.in/gwMNu-fC

Quantum Neural network on a Hybrid-Classical Computer with: num_layers = 3 num_neurons = 10 num_qubits = 5 num_classical_neurons = 20

Simple neural network on STM32.

Still experiencing spurious alarms from your flame detector from modulated sunlight or black body radiation? Explore our new General Monitors® FL5000 MSIR flame with Artificial Neural Network (ANN) technology to understand how our combined technology will help you focus on your operations. 👉 Learn the difference today: https://lnkd.in/gpKsQNZ3 #MSAsafety #FGFD #FlameDetector #SafeguardingPeoplePlacesandThePlanet

Introducing the brand new Neuromorphic Neural Engine Module NNE-5508. This module features fast computing with pattern recognition abilities by deploying the neuromorphic technology of General Vision.

Still experiencing spurious alarms from your flame detector from modulated sunlight or black body radiation? Explore our new General Monitors® FL5000 MSIR flame with Artificial Neural Network (ANN) technology to understand how our combined technology will help you focus on your operations. 👉 Learn the difference today: https://lnkd.in/g2gu7HqS #MSAsafety #FGFD #FlameDetector #SafeguardingPeoplePlacesandThePlanet

Deep neural network based models for radiowave propagation take time to train. This time pays off, if the models can be applied (a) to new geometries and (b) to tasks that require repetitive simulations, such as the optimal selection of transmitter locations. Aris Seretis, Charley X. and I demonstrate (in simulation and experiment) a neural net propagation model that dramatically accelerates the selection of wireless access point locations (on IEEE Transactions on Antennas and Propagation early access): https://lnkd.in/g47B2iMX #wireless #aiml #propagation #nserc

Hi everyone! Exciting news: the TensorKrowch paper has finally been published! 🎉 https://lnkd.in/d6E5Jjzg I know it's been a while since my last update, so let me share the latest advancements in the library: The latest releases have introduced several powerful functionalities to tensorize neural networks, train generative TN models, interpret TNs through reduced densities and entanglement entropies, and tackle physics-inspired problems like energy minimization. - MPS classes have been rebuilt, keeping the original functionality almost intact, but adding many new possibilities. One very useful feature is that now you can select which sites of the MPS are for input/output, and marginalize the output sites upon contraction. - This makes it straightforward to compute marginal distributions, reduced densities and norms. In fact, all these functionalities have also been added as methods of MPS models. Besides, these additions enable to train generative models out of the box. - Also, with the option to marginalize output nodes and the support for complex numbers, you can use TensorKrowch for energy minimization of MPS models, given a Hamiltonian in MPO form. - MPS methods now have the option to renormalize, which allows to train models with more flexible initializations, still getting manageable results. - There are also new initializations and embeddings, so you can try the combination that best fits your problem for training. - New models have been added, like MPO and MPSData. These models, together with the new decompositions vec_to_mps and mat_to_mpo, allow for the seamless tensorization of neural networks. - There is also a new page of examples in the documentation, where you can see how to train different TN models. It includes training MPS models (with and without DMRG), hybrid tensorial neural network models, tensorization of neural networks, and more to come! One of the goals of TensorKrowch is to allow researchers to easily reproduce results/experiments from papers. If you make your own implementations, I encourage you to contribute to the project by adding them to the examples page! I am very happy with the stage TensorKrowch is in right now. It's a friendly tool with which you can explore most of the applications of TNs in machine learning and even use it for physics applications! I hope you find it useful as well! 😊

Design of a unified physics-informed neural network using interior point algorithm to study the bioconvection nanofluid flow via stretching surface