Debapam Pal’s Post

I am pleased to announce that our paper, "Imagination Made Real: Stable Diffusion for High-Fidelity Text-to-Image Tasks" has been published in IEEE Xplore! In this work, we explored the power of diffusion models (DMs) for high-quality text-to-image generation, overcoming the limitations of traditional methods that demand heavy computational resources. By leveraging latent spaces of pretrained autoencoders, we were able to significantly reduce the computational burden while maintaining exceptional visual fidelity. Our approach also integrates cross-attention layers, turning DMs into versatile generators capable of handling various conditioning inputs like text and bounding boxes. This research showcases the potential of latent diffusion models (LDMs) to outperform traditional pixel-based DMs across various tasks, including image inpainting, class-specific image blending, and more. I, along with my team consisting of Manas jain, Devika Kadam, Harshvardhan Kadam and Toshish Kakkad have presented this paper at the 2024 2nd International Conference on Sustainable Computing and Smart Systems (ICSCSS 2024). You can check out the paper here: https://lnkd.in/d7iaKTpe #Research #TextToImage #MachineLearning #AI #DeepLearning #IEEE

A reminder of Prof. Iskander's lecture next week on bio-inspired computational EM and artificial intelligence! You must register in advance in order to attend. https://lnkd.in/gE-BFNBM

Are you passionate about AI-driven molecular communication and its transformative potential in medicine? 🔬🧠 Join upcoming IEEE ICMLCN Conference 2025 for our special session dedicated to exploring how Machine Learning/Deep Learning is shaping the future of the Internet of Bio-Nano Things (IoBNT). AI plays a critical role in this transformation by decoding complex molecular signals, optimizing drug delivery, and guiding nano-robots to perform life-saving tasks with high precision. From early disease detection to personalized treatments tailored to individual patients, this fusion of AI and bio-nano technologies is opening up exciting possibilities in healthcare. Here is the link: https://lnkd.in/eysZw7Az #AI #MolecularCommunication #Medicine #Healthcare

Very happy to introduce the recently published articles from ACM Transactions on Probabilistic Machine Learning (TOPML), a newly launched Gold OA journal focusing on probabilistic methods that learn from data to improve performance on decision-making or prediction tasks under uncertainty. DRD-GAN: A Novel Distributed Conditional Wasserstein Deep Convolutional Relativistic Discriminator GAN with Improved Convergence Arunava Roy, PhD, Dipankar Dasgupta Stochastic Optimization and Learning for Two-Stage Supplier Problems Brian Brubach, Nathaniel Grammel, David Harris, Aravind Srinivasan, Leonidas Tsepenekas, Anil Kumar Vullikanti Elliptically-Contoured Tensor-variate Distributions with Application to Image Learning Carlos Llosa, Ranjan Maitra https://lnkd.in/erDRmQjx #machinelearning #probabilisticmachinelearning #optimization Theodore Papamarkou, Fang Liu, Wray Buntine, ACM Digital Library

In June, IQ will be found in the historical heart of modern civilization, on the island of Crete! More precisely in historic Fodele, 26 minutes outside of Heraklion. IQ Senior Scientist Dr. Alexander Katrompas has been accepted and invited to present his latest research paper "Many-to-Many Prediction for Effective Modeling of Frequent Label Transitions in Time Series" during PETRA 2024. PETRA is part of the Association for Computing Machinery International Conference Proceeding Series. From the abstract: "... this paper presents a novel many-to-many time-series model and post-processing using hybrid recurrent neural networks with attention mechanisms, which more effectively captures label transitions over traditional many-to-one models. Further, unlike typical other many-to-many models, our approach doesn't require a decoder. Instead, it employs an RNN, generating a label for every input time step. During inference, a weighted voting scheme consolidates overlapping predictions into one label per time step. Experiments show our model remains effective on time-series with sparse label shifts, but particularly excels in detecting frequent transitions. This model is ideal for tasks demanding accurate pinpointing of rapid label changes in time-series data... " Dr. Katrompas's novel research again represents the leading edge of time-series analytics for asset and process intensive applications, such as applied Scientific Machine Learning for Heavy Industries (Energy, Oil & Gas, Petrochemicals, Chemicals, and more). All publishing and reproduction rights reserved by the ACM Digital Library as part of the ACM ICPS program. #txst #utarlington #nsf #sciml #ml #heavyindustry

Congratulations to my co-authors, Hao Zhou and Melike Erol-Kantarci , on the selection of our paper “Learning from Peers: Deep Transfer Reinforcement Learning for Joint Radio and Cache Resource Allocation in 5G RAN Slicing” which appeared in the IEEE Transactions on Cognitive Communications and Networking, Dec. 2022, pp. 1925-1941, for the IEEE ComSoc CSIM TC Best Journal Paper Award. This award will be presented at the upcoming IEEE International Conference on Communications, to be held in Denver, CO, in June 2024. This paper can be found at: https://lnkd.in/eA5a3C2T This work investigates how transfer learning can improve resource slicing at the 5G edge, aiming to enhance the training efficiency of AI-enabled 5G RAN. It proposed two deep transfer reinforcement learning algorithms and revealed two knowledge transfer approaches from expert agents to learner agents. A comprehensive overview of knowledge transfer in RAN slicing can be found in our companion magazine paper, Knowledge Transfer and Reuse: A Case Study of AI-Enabled Resource Management in RAN Slicing,” IEEE Wireless Communications, Oct. 2023, vol.30. no.5, which can be found at:   https://lnkd.in/enA8ayqx  

I'm very excited to share that our paper titled "Error Detection and Correction Codes for Safe In-Memory Computations" has been accepted as a poster at the European Test Symposium (ETS) to be held in The Hague! I would like to thank my supervisors and the co-authors who worked with me on this project: Taha Soliman, Benjamin Hettwer, Jan Micha Borrmann, Simranjeet Singh, Ankit Bende, Dr. Vikas Rana, Farhad Merchant and Norbert Wehn. The paper delves into the crucial aspects of safety for In-Memory Computing architectures applied to the AI domain. We show how it is possible to recover the baseline accuracy of multiple algorithms, by detecting and correcting errors at runtime through the coordinated action of multiple checksum codes. Finally, we illustrate the trade-off in terms of area and latency overhead required by such architectural mitigation technique and we compare it with Triple Modular Redundancy (TMR) and other state-of-the-art solutions. The pre-print is available at: https://lnkd.in/dUVDt56E #IMC #AI #Safety #Checksum #ETS #BoschResearch #Publication #arXiv

How to submit | Machine learning for materials discovery and optimization 2024 - Nature: ... artificial intelligence and high-throughput approaches. Advances in machine learning for materials science, data-driven materials prediction ... http://dlvr.it/TFY2gy

The right direction

#Publication Update ! I am thrilled to share that our paper has been accepted and was recently presented at IEEE International Conference on Artificial Intelligence and Machine Learning Applications(AIMLA 2024). I would like to extend my heartfelt gratitude to my co-authors, Yash Shingavi and Professor Washima Tasnin, for their unwavering support and invaluable contributions throughout this journey. Soon you can find this paper on IEEE Xplore.  #EnsembleLearning #MachineLearning #AIConference #IEEE #Research #Innovation