This work investigates an important phenomenon in centroid-based deep clustering (DC) algorithms: Performance quickly saturates after a period of rapid early gains. Practitioners commonly address early saturation with periodic reclustering, which we demonstrate to be insufficient to address performance plateaus. We call this phenomenon the “reclustering barrier” and empirically show when the reclustering barrier occurs, what its underlying mechanisms are, and how it is possible to Break the Reclustering Barrier with our algorithm BRB. BRB avoids early over-commitment to initial clusterings and enables continuous adaptation to reinitialized clustering targets while remaining conceptually simple. Applying our algorithm to widely-used centroid-based DC algorithms, we show that (1) BRB consistently improves performance across a wide range of clustering benchmarks, (2) BRB enables training from scratch, and (3) BRB performs competitively against state-of-the-art DC algorithms when combined with a contrastive loss. We release our code and pre-trained models at https://github.com/Probabilistic-and-Interactive-ML/breaking-the-reclustering-barrier .

We present Cell2Cell, a novel visual analytics approach for quantifying and visualizing networks of cell-cell interactions in three-dimensional (3D) multi-channel cancerous tissue data. By analyzing cellular interactions, biomedical experts can gain a more accurate understanding of the intricate relationships between cancer and immune cells. Recent methods have focused on inferring interaction based on the proximity of cells in low-resolution 2D multi-channel imaging data. By contrast, we analyze cell interactions by quantifying the presence and levels of specific proteins within a tissue sample (protein expressions) extracted from high-resolution 3D multi-channel volume data. Such analyses have a strong exploratory nature and require a tight integration of domain experts in the analysis loop to leverage their deep knowledge. We propose two complementary semi-automated approaches to cope with the increasing size and complexity of the data interactively: On the one hand, we interpret cell-to-cell interactions as edges in a cell graph and analyze the image signal (protein expressions) along those edges, using spatial as well as abstract visualizations. Complementary, we propose a cell-centered approach, enabling scientists to visually analyze polarized distributions of proteins in three dimensions, which also captures neighboring cells with biochemical and cell biological consequences. We evaluate our application in three case studies, where biologists and medical experts use Cell2Cell to investigate tumor micro-environments to identify and quantify T-cell activation in human tissue data. We confirmed that our tool can fully solve the use cases and enables a streamlined and detailed analysis of cell-cell interactions.

Andreas Stephan, Lukas Miklautz, Kevin Sidak, Jan Philip Wahle, Bela Gipp, Claudia Plant, Benjamin Roth. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Quantizing deep neural networks (DNNs) is an important strategy for training or inference in time critical applications. State-of-the-art quantization approaches focus on post-training quantization. While some work on quantization during training exists, most approaches require refinement in full precision (usually single precision) in the final training phase, use a rather coarse quantization, that leads to a loss in accuracy, or enforce a global bit-width across the entire DNN. This leads to suboptimal assignments of bit-widths to layers and, consequently, suboptimal resource usage. To overcome such limitations, we introduce AdaPT, a new fixed- point quantized sparsifying training strategy for deep neural networks. AdaPT decides about precision switches between training epochs based on an information theory motivated heuristic. On a per-layer basis, AdaPT chooses the lowest precision that causes no quantization-induced information loss, while keeping the precision high enough such that future learning steps do not suffer from vanishing gradients. The benefits of this quantization are evaluated based on an analytical performance model. We illustrate an average 1.31 × (or 4.76× adjusted for iso-accuracy) speedup compared to standard training in float32 at iso-accuracy, even achieving an average accuracy increase of 0.74 percentage points for AlexNet/ResNet-20 on CIFAR10/CIFAR100/EMNIST and LeNet-5/MNIST. We demonstrate that these trained models reach an average inference 2.28× speedup with a model size reduction up to 51% of the corresponding unquantized model.