The Platonic Representation Hypothesis claims that recent foundation models are converging to a shared representation space as a function of their downstream task performance, irrespective of the objectives and data modalities used to train these models (Huh et al., 2024). Representational similarity is generally measured for individual datasets and is not necessarily consistent across datasets. Thus, one may wonder whether this convergence of model representations is confounded by the datasets commonly used in machine learning. Here, we propose a systematic way to measure how representational similarity between models varies with the set of stimuli used to construct the representations. We find that the objective function is a crucial factor in determining the consistency of representational similarities across datasets. Specifically, self-supervised vision models learn representations whose relative pairwise similarities generalize better from one dataset to another compared to those of image classification or image-text models. Moreover, the correspondence between representational similarities and the models’ task behavior is dataset-dependent, being most strongly pronounced for single-domain datasets. Our work provides a framework for analyzing similarities of model representations across datasets and linking those similarities to differences in task behavior.

Deep neural networks have reached human-level performance on many computer vision tasks. However, the objectives used to train these networks enforce only that similar images are embedded at similar locations in the representation space, and do not directly constrain the global structure of the resulting space. Here, we explore the impact of supervising this global structure by linearly aligning it with human similarity judgments. We find that a naive approach leads to large changes in local representational structure that harm downstream performance. Thus, we propose a novel method that aligns the global structure of representations while preserving their local structure. This global-local transform considerably improves accuracy across a variety of few-shot learning and anomaly detection tasks. Our results indicate that human visual representations are globally organized in a way that facilitates learning from few examples, and incorporating this global structure into neural network representations improves performance on downstream tasks.

Estimating what would be an individual's potential response to varying levels of exposure to a treatment is of high practical relevance for several important fields, such as healthcare, economics and public policy. However, existing methods for learning to estimate counterfactual outcomes from observational data are either focused on estimating average dose-response curves, or limited to settings with only two treatments that do not have an associated dosage parameter. Here, we present a novel machine-learning approach towards learning counterfactual representations for estimating individual dose-response curves for any number of treatments with continuous dosage parameters with neural networks. Building on the established potential outcomes framework, we introduce performance metrics, model selection criteria, model architectures, and open benchmarks for estimating individual dose-response curves. Our experiments show that the methods developed in this work set a new state-of-the-art in estimating individual dose-response.

Balancing accuracy gains with other objectives such as interpretability is a key challenge when building decision trees. However, this process is difficult to automate because it involves know-how about the domain as well as the purpose of the model. This paper presents TreePOD, a new approach for sensitivity-aware model selection along trade-offs. TreePOD is based on exploring a large set of candidate trees generated by sampling the parameters of tree construction algorithms. Based on this set, visualizations of quantitative and qualitative tree aspects provide a comprehensive overview of possible tree characteristics. Along trade-offs between two objectives, TreePOD provides efficient selection guidance by focusing on Pareto-optimal tree candidates. TreePOD also conveys the sensitivities of tree characteristics on variations of selected parameters by extending the tree generation process with a full-factorial sampling. We demonstrate how TreePOD supports a variety of tasks involved in decision tree selection and describe its integration in a holistic workflow for building and selecting decision trees. For evaluation, we illustrate a case study for predicting critical power grid states, and we report qualitative feedback from domain experts in the energy sector. This feedback suggests that TreePOD enables users with and without statistical background a confident and efficient identification of suitable decision trees.