Dataset Information

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets.

ABSTRACT: Recently, deep learning has unlocked unprecedented success in various domains, especially using images, text, and speech. However, deep learning is only beneficial if the data have nonlinear relationships and if they are exploitable at available sample sizes. We systematically profiled the performance of deep, kernel, and linear models as a function of sample size on UKBiobank brain images against established machine learning references. On MNIST and Zalando Fashion, prediction accuracy consistently improves when escalating from linear models to shallow-nonlinear models, and further improves with deep-nonlinear models. In contrast, using structural or functional brain scans, simple linear models perform on par with more complex, highly parameterized models in age/sex prediction across increasing sample sizes. In sum, linear models keep improving as the sample size approaches ~10,000 subjects. Yet, nonlinearities for predicting common phenotypes from typical brain scans remain largely inaccessible to the examined kernel and deep learning methods.

SUBMITTER: Schulz MA

PROVIDER: S-EPMC7447816 | biostudies-literature | 2020 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets.

Schulz Marc-Andre MA Yeo B T Thomas BTT Vogelstein Joshua T JT Mourao-Miranada Janaina J Kather Jakob N JN Kording Konrad K Richards Blake B Bzdok Danilo D

Nature communications 20200825 1

Recently, deep learning has unlocked unprecedented success in various domains, especially using images, text, and speech. However, deep learning is only beneficial if the data have nonlinear relationships and if they are exploitable at available sample sizes. We systematically profiled the performance of deep, kernel, and linear models as a function of sample size on UKBiobank brain images against established machine learning references. On MNIST and Zalando Fashion, prediction accuracy consiste ...[more]

PMID: 32843633

Similar Datasets

Project description:In highly non-linear datasets, attributes or features do not allow readily finding visual patterns for identifying common underlying behaviors. Therefore, it is not possible to achieve classification or regression using linear or mildly non-linear hyperspace partition functions. Hence, supervised learning models based on the application of most existing algorithms are limited, and their performance metrics are low. Linear transformations of variables, such as principal components analysis, cannot avoid the problem, and even models based on artificial neural networks and deep learning are unable to improve the metrics. Sometimes, even when features allow classification or regression in reported cases, performance metrics of supervised learning algorithms remain unsatisfyingly low. This problem is recurrent in many areas of study as, per example, the clinical, biotechnological, and protein engineering areas, where many of the attributes are correlated in an unknown and very non-linear fashion or are categorical and difficult to relate to a target response variable. In such areas, being able to create predictive models would dramatically impact the quality of their outcomes, generating an immediate added value for both the scientific and general public. In this manuscript, we present RV-Clustering, a library of unsupervised learning algorithms, and a new methodology designed to find optimum partitions within highly non-linear datasets that allow deconvoluting variables and notoriously improving performance metrics in supervised learning classification or regression models. The partitions obtained are statistically cross-validated, ensuring correct representativity and no over-fitting. We have successfully tested RV-Clustering in several highly non-linear datasets with different origins. The approach herein proposed has generated classification and regression models with high-performance metrics, which further supports its ability to generate predictive models for highly non-linear datasets. Advantageously, the method does not require significant human input, which guarantees a higher usability in the biological, biomedical, and protein engineering community with no specific knowledge in the machine learning area.

Project description:This paper presents datasets utilised for synthetic near-infrared (NIR) image generation and bounding-box level fruit detection systems. A high-quality dataset is one of the essential building blocks that can lead to success in model generalisation and the deployment of data-driven deep neural networks. In particular, synthetic data generation tasks often require more training samples than other supervised approaches. Therefore, in this paper, we share the NIR+RGB datasets that are re-processed from two public datasets (i.e., nirscene and SEN12MS), expanded our previous study, deepFruits, and our novel NIR+RGB sweet pepper (capsicum) dataset. We oversampled from the original nirscene dataset at 10, 100, 200, and 400 ratios that yielded a total of 127 k pairs of images. From the SEN12MS satellite multispectral dataset, we selected Summer (45 k) and All seasons (180k) subsets and applied a simple yet important conversion: digital number (DN) to pixel value conversion followed by image standardisation. Our sweet pepper dataset consists of 1615 pairs of NIR+RGB images that were collected from commercial farms. We quantitatively and qualitatively demonstrate that these NIR+RGB datasets are sufficient to be used for synthetic NIR image generation. We achieved Frechet inception distances (FIDs) of 11.36, 26.53, and 40.15 for nirscene1, SEN12MS, and sweet pepper datasets, respectively. In addition, we release manual annotations of 11 fruit bounding boxes that can be exported in various formats using cloud service. Four newly added fruits (blueberry, cherry, kiwi and wheat) compound 11 novel bounding box datasets on top of our previous work presented in the deepFruits project (apple, avocado, capsicum, mango, orange, rockmelon and strawberry). The total number of bounding box instances of the dataset is 162 k and it is ready to use from a cloud service. For the evaluation of the dataset, Yolov5 single stage detector is exploited and reported impressive mean-average-precision, mAP[0.5:0.95] results of min:0.49, max:0.812. We hope these datasets are useful and serve as a baseline for future studies.

Dataset Information

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets.

Publications

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets