In this episode, we explore the nuances and key considerations of implementing graph neural networks as decentralized applications in wireless networks, such as source localization, multi-robot flocking, and wireless channel management — a core theme of this podcast, especially in this season. This discussion is based on a journal paper published in IEEE Transactions on Signal Processing, authored by Zhan Gao and Deniz Gündüz.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the paper here: [IEEE TSP](https://ieeexplore.ieee.org/document/9042352)

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into the applications of graph neural networks as a learnable digital twin of network simulators, which can accelerate network optimization by its fast and differentiable prediction of networking key performance indicators (KPIs). This episode is based on a preprient authored by Boning Li, et al.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the paper here: [arXiv]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into a survey paper of the applications of graph neural networks in electrical engineering. This episode is based on our recent publication in Nature Review Electrical Engineering.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the paper here: [Online]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into fully distributed online training of graph neural networks deployed in networked systems. This episode is based on our recent preprint by Rostyslav Olshevskyi from Rice University.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the paper here: [Preprint]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into the combination of graph neural networks and unsupervised primal-dual learning, a model-free approach to scalable, intelligent wireless resource allocations. This episode is based on two papers published in IEEE Transactions on Signal Processing by Zhiyang Wang, Mark Eisen, and Alejandro Ribeiro.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the papers here: [Eisen 2019], [Wang 2022]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into annealed Langevin dynamics, a new approach to linear inverse problems, such as massive MIMO detection, image deblurring, and network topology inference. This episode is based on two papers published in IEEE Transactions on Wireless Communications (TWC) and IEEE Transactions on Signal Processing by Nicolas Zilberstein, et. al.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the papers here: [TWC], [TSP]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into neural OFDM receivers, a series of four papers published in IEEE Journal on Selected Areas in Communications (JSAC) and IEEE Transactions on Wireless Communications (TWC) on the topic of leveraging deep learning to enhance radio signal reception in Orthogonal Frequency-Division Multiplexing (OFDM) systems.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.

🎧 Read the papers here: [Deep-Waveform], [DeepRx], [Power Efficiency] [Pilotless Communications]

📷 Cover Image Source: imagine.art, Microsoft Designer

🎵 BGM: Artlist.io

🛠️ Credits: NotebookLM by Google

In this episode, we dive into our own tutorial (IEEE ICMLCN, MILCOM 2024) on how could graph neural networks be used to enhance wireless communications.

Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.
🎧 Read the slides here: [Link to tutorial]
📷 Cover Image Source: imagine.art
🎵 BGM: Artlist.io
🛠️ Credits: NotebookLM by Google

In this episode, we dive into distributional reinforcement learning, including three papers from Google DeepMind and another Cell paper from Harvard. Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research topic. NotebookLM can make mistakes, so double-check it.
🎧 Read the papers here: [C51, QR-DQN, IQN, Cell paper]
📷 Cover Image Source: imagine.art
🎵 BGM: Artlist.io
🛠️ Credits: NotebookLM by Google

Description: In this episode, we dive into our own paper published in ICLR 2023, a study on scalable reinforcement learning for networked systems. Generated using NotebookLM from Google, this podcast highlights the key findings and implications of this research.
🎧 Read the paper here: [openreview] [video presentation]
📷 Cover Image Source: imagine.art
🎵 BGM: Artlist.io
🛠️ Generated by: NotebookLM by Google