Position: Embracing Negative Results in Machine Learning

Authors: Florian Karl, Malte Kemeter, Gabriel Dax, Paulina Sierak

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this position paper we argue that predictive performance alone is not a good indicator for the worth of a publication. Using it as such even fosters problems like inefficiencies of the machine learning research community as a whole and setting wrong incentives for researchers.
Researcher Affiliation Collaboration 1Fraunhofer Institute for Integrated Circuits IIS, Fraunhofer IIS, Nuremberg, Germany 2Ludwig-Maximilians-Universit at M unchen, Munich, Germany 3Munich Center for Machine Learning, Munich, Germany.
Pseudocode No The paper is a position paper discussing the importance of negative results in machine learning and does not propose any new algorithms or methods that would require pseudocode or algorithm blocks.
Open Source Code No The paper is a position paper and does not present novel methodology that would have associated source code. It discusses the concept of open-source code in the context of research practices but does not provide any for its own content.
Open Datasets No The paper is a position paper and does not conduct experiments. It refers to established datasets like ImageNet or CIFAR as examples in the discussion of machine learning practices, but these are not used for empirical work within the paper itself.
Dataset Splits No The paper is a position paper and does not conduct experiments, thus it does not provide details on training, validation, or test dataset splits.
Hardware Specification No The paper is a position paper and does not conduct experiments, thus it does not specify hardware used for running experiments. It generally discusses the role of computing resources in the field but not its own usage.
Software Dependencies No The paper is a position paper and does not describe a method that requires specific software dependencies with version numbers for replication.
Experiment Setup No The paper is a position paper and does not conduct experiments, therefore it does not provide details about an experimental setup or hyperparameters.