2,670,063 research outputs found
Metadata Representations for Queryable ML Model Zoos
No full textMachine learning (ML) practitioners and organizations are building model zoos of pre-trained models, containing metadata describing properties of the ML models and datasets that are useful for reporting, auditing, reproducibility, and interpretability purposes. The metatada is currently not standardised; its expressivity is limited; and there is no interoperable way to store and query it. Consequently, model search, reuse, comparison, and composition are hindered. In this paper, we advocate for standardized ML model metadata representation and management, proposing a toolkit supported to help practitioners manage and query that metadata.Web Information SystemsHuman-Centred Artificial Intelligenc
A Manifesto of Nodalism
Get PDFThis paper proposes the notion of Nodalism as a means describing contemporary culture and of understanding my own creative practice in electronic music composition. It draws on theories and ideas from Kirby, Bauman, Bourriaud, Deleuze, Guatarri, and Gochenour, to demonstrate how networks of ideas or connectionist neural models of cognitive behaviour can be used to contextualize, understand and become a creative tool for the creation of contemporary electronic music
A Simple Nested Simulation for SED-ML
Get PDFThis document describes a simple nested Simulation Experiment for SED-ML [1] that is easy to implement and will help to broaden what SED-ML is able to encode
Modular-ML-inference/ml-usecase: v1.0.0
No full textUse case preparation files, including test data, ML models, and data preprocessing notebooks
Modular-ML-inference/ml-usecase: v0.1.0
No full textUse case preparation files, including test data, ML models, and data preprocessing notebooks
Optimizing ML Inference Queries Under Constraints
No full textThe proliferation of pre-trained ML models in public Web-based model zoos facilitates the engineering of ML pipelines to address complex inference queries over datasets and streams of unstructured content. Constructing optimal plan for a query is hard, especially when constraints (e.g. accuracy or execution time) must be taken into consideration, and the complexity of the inference query increases. To address this issue, we propose a method for optimizing ML inference queries that selects the most suitable ML models to use, as well as the order in which those models are executed. We formally define the constraint-based ML inference query optimization problem, formulate it as a Mixed Integer Programming (MIP) problem, and develop an optimizer that maximizes accuracy given constraints. This optimizer is capable of navigating a large search space to identify optimal query plans on various model zoos.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Web Information SystemsHuman-Centred Artificial Intelligenc
BiomedSciAI/fuse-med-ml: 0.2.4
No full textWhat's Changed
Add supported python versions badge by @SagiPolaczek in https://github.com/BiomedSciAI/fuse-med-ml/pull/183
multimodal modifications by @mosheraboh in https://github.com/BiomedSciAI/fuse-med-ml/pull/184
Full Changelog: https://github.com/BiomedSciAI/fuse-med-ml/compare/0.2.3...0.2.
Parfait-ML-1.0.0.zip
No full textParfait-ML is a search-based software testing and statistical debugging tool to configure ML libraries fairly and detect fairness bugs in the configuration space of ML algorithms. Parfait-ML focuses on the hyperparameters in training of ML models and whether they amplify or suppress discriminations in data-driven software. Some prominent examples of hyperparameters include l1 vs. l2 loss function in support vector machines, the maximum depth of a decision tree, and the number of layers/neurons in deep neural networks.Parfait-ML consists of two components: 1) search algorithms 2) data-driven explanations. The tool has three search algorithms to explore the space of ML hyperparameters in order to find what hyperparameter values can lead to high and low fairness. The current experiments include five ML algorithms and their hyperparameter spaces (e.g., logistic regression). To train ML algorithms, Parfait-ML includes four social-critical datasets (e.g., adult census income). During the search, the tool consider the default configuration of ML algorithm and its performance as a lower-bound on the functional accuracy requirements. Then, it selects inputs that satisfy the accuracy requirements and lead to an interesting fairness value. In addition, the coverage-based algorithm (graybox search) uses feedback from program internals to identify interesting inputs. For the fairness metric, we use EOD: the difference in true positive rates between two groups, and AOD: the average difference between true positive rates and false positive rates. We note that these metrics measure biases and higher values mean lower fairness.After stopping the search algorithm (a default of 4 hours), we use clustering algorithms to partition the hyperparameter inputs generated from the search based on fairness and accuracy values. Given two groups of inputs with low and high biases and similar accuracy, we infer CART decision trees that show what hyperparameters are common inside a cluster and what hyperparameters distinguish low fairness configurations from the high fairness one.</p
- …
