TECHNICAL PROGRAM
Tentative Program Overview
Note that the program below is tentative, and is subject to change. It can however give you a general idea for the time being. We will try to release the detailed program before August.

All papers will do a poster presentation (regular track, special session track, data competition track, ICASSP 2025 papers, SPS papers).
We have selected a small number of papers for oral presentations. Here are the selected papers:
Oral Session 1 papers: Signal Decomposition and Estimation
- (regular track) Continuous-Time Signal Decomposition: An Implicit Neural Generalization of PCA and
ICA
- (regular track) Learning Rate Should Scale Inversely with High-Order Data Moments in High-Dimensional Online Independent Component Analysis
- (regular track) LANM: Learned Atomic Norm Minimization for Superfast Gridless Spectral Compressed Sensing
- (regular track) prNet: Efficient and Robust Phase Retrieval via Stochastic Refinement
Coordinate Ascent Neural Kalman-MLE for State Estimation
- (regular track) Coordinate Ascent Neural Kalman-MLE for State Estimation
Oral Session 2 papers: Audio, Speech, and Music Processing
- (Data competition: NOCASA) Comparison of End-to-end Speech Assessment Models for the NOCASA 2025 Challenge
- (regular track) Audio Prototypical Network for Controllable Music Recommendation
- (regular track) Re-Bottleneck: Latent Re-Structuring for Neural Audio Autoencoders
- (regular track) DDL: A Dataset for Drone Detection and Localization from Multi-Channel Audio and a Deep Uncertainty-Aware Framework
- (regular track) Input Conditioned Layer Dropping in Speech Foundation Models
Oral Session 3 papers: Learning Algorithms and Optimization
- (Special Session: LEAP) Improving Communication-Efficiency for Decentralized Federated Clustering
- (regular track) Fast and Robust Training of Deep Learning Models with Multiplicative Adagrad
- (regular track) Information Entropy-Based Scheduling for Communication-Efficient Decentralized Learning
- (regular track) Model Recycling Framework for Multi-Source Data-Free Supervised Transfer Learning
- (regular track) Meta-Tree: Bayesian Approach to Avoid Overfitting in Decision Trees and Analysis on the Application to Boosting
Oral Session 4 papers: Computer Vision
- (Special Session: LVLM) Trust the Model: Compact VLMS As In-Context Judges For Image-test Data Quality
- (Special Session: Art) Solving Jigsaw Puzzles in the Wild: Human-Guided Reconstruction of Cultural Heritage Fragments
- (regular track) APA: Domain Generalization Using Frequency Based Augmentation
- (regular track) Ptychographic Image Reconstruction from Limited Data via Score-Based Diffusion Models with Physics-Guidance
- (regular track) RadioTrace: Bridging Diffusion Priors and RSS Measurements for Accurate Radio Map Estimation
Oral Session 5 papers: ML for Health and Neuroscience
- (Special Session: Brain Decoding) Toward a gaze-independent brain-computer interface using the code-modulated visual evoked potentials
- (regular track) Closing the Gap in Multimodal Medical Representation Alignment
- (regular track) Cycle-Consistent Diffusion Model with Vessel-Aware Attention for Endoscopic Image Translation
- (regular track) RSR-NF: Neural Field Regularization by Static Restoration Priors for Computed Dynamic Imaging
- (regular track) Perturbation-based Multiview Graph Learning with Consensus Graph