end-to-end optimization
This technique refers to the process of optimizing an entire model directly from input to output, often through backpropagation, allowing for more integrated and potentially more effective learning as all components of the model are adjusted collectively.
- Gaussian-Augmented Physics Simulation and System Identification with Complex Colliders
- Information-Driven Design of Imaging Systems
- LinPrim: Linear Primitives for Differentiable Volumetric Rendering
- PermLLM: Learnable Channel Permutation for N:M Sparse Large Language Models
- Real-Time Scene-Adaptive Tone Mapping for High-Dynamic Range Object Detection
- Revisiting End-to-End Learning with Slide-level Supervision in Computational Pathology
- TaDiCodec: Text-aware Diffusion Speech Tokenizer for Speech Language Modeling
- Týr-the-Pruner: Structural Pruning LLMs via Global Sparsity Distribution Optimization