About
Research & Development
ScenGAN: Enhancing Renewable Energy Forecasting with Advanced Uncertainty Modeling
TLDR: ScenGAN is a deep generative model that uses attention mechanisms and Bayesian deep learning to forecast renewable energy scenarios. It effectively captures both...
Research & Development
MaskVCT: A New Approach to Controllable Zero-Shot Voice Conversion
TLDR: MaskVCT is a new zero-shot voice conversion model that offers unprecedented multi-factor controllability. It uses masked generative modeling, syllabic representations, and multiple classifier-free...
Research & Development
Enhancing 3D Point Cloud Models with Replaced Token Denoising
TLDR: Point-RTD is a novel pretraining strategy for transformer models on 3D point clouds. Unlike traditional masked autoencoding, it corrupts point cloud tokens and...
Research & Development
Neural Network Safety: Addressing Tail Risks with Worst-Case CVaR
TLDR: This research introduces a new framework for verifying the safety of neural networks by integrating Worst-Case Conditional Value-at-Risk (WC-CVaR) into existing quadratic-constraint and...
Research & Development
Autiverse: Empowering Autistic Adolescents to Share Their Stories Through AI Journaling
TLDR: Autiverse is an AI-guided multimodal journaling app designed for autistic adolescents to improve narrative skills and emotional expression. It uses conversational AI and...
Research & Development
Enhancing Data Integrity: A Hybrid Approach to Entity Resolution for Enterprise Systems
TLDR: The "Transformer-Gather, Fuzzy-Reconsider" (TGFR) framework is a scalable hybrid solution for entity resolution. It combines deep learning (transformers for semantic understanding and efficient...