Code release for Proto-CLIP: Vision-Language Prototypical Network for Few-Shot Learning | IROS 2024

A Lightweight and High Performance Neural network for MI-EEG decoding

Case-based interpretable deep learning for ECG classification. This code implements ProtoECGNet from the following paper: "ProtoECGNet: Case-Based Interpretable Deep Learning for Multi-Label ECG Classification." Sethi et al., MLHC 2025

Hyperbolic Busemann Learning with Ideal Prototypes, NeurIPS2021

Official Pytorch Code of Our Paper: Rethinking Multiple Instance Learning for Whole Slide Image Classification: A Good Instance Classifier is All You Need

[WACV 2025] LatteCLIP: Unsupervised CLIP Fine-Tuning via LMM-Synthetic Texts

[MM2024] PyTorch implementation of "Semantic Codebook Learning for Dynamic Recommendation Models".

[CIKM 2024] PRISM: Mitigating EHR Data Sparsity via Learning from Missing Feature Calibrated Prototype Patient Representations

Case-based interpretable deep learning for ECG classification. This code implements ProtoECGNet from the following paper: "ProtoECGNet: Case-Based Interpretable Deep Learning for Multi-Label ECG Classification." Sethi et al., MLHC 2025

Source code for the ADAM submission on the PlantCLEF2025 challenge

Does the Tversky projection layer's gain come from asymmetric similarity or its feature bank? A controlled decomposition on frozen self-supervised ImageNet-100 embeddings. Finding: on coarse classes neither beats a linear baseline, and richer metrics make class prototypes less central and less stable.

Frozen encoder + Mahalanobis prototype for class-incremental intent classification. 50+ experiments across BANKING77, CLINC150, HWU64, AG News. Matches fine-tuned baselines at 5MB state with zero forgetting, order-invariance, 455 QPS.

Existing cross-domain recommendation methods rely on overlapping users and source interaction sequences to learn preference transfer function, leaving the rich structural information in pre-trained target-domain embeddings unexploited. In this paper, we propose COTA (Cluster Optimal Transport Alignment), a framework for cold-start CDR.

A lightweight Few-Shot Image Classifier built with PyTorch, capable of recognizing new classes from just a few examples using prototype-based learning. Designed for easy model export and integration.