Human-in-the-loop optimization of visual prosthetic stimulation DOI
Tristan Fauvel, Matthew Chalk

Journal of Neural Engineering, Год журнала: 2022, Номер 19(3), С. 036038 - 036038

Опубликована: Июнь 1, 2022

Abstract Objective . Retinal prostheses are a promising strategy to restore sight patients with retinal degenerative diseases. These devices compensate for the loss of photoreceptors by electrically stimulating neurons in retina. Currently, visual function that can be recovered such is very limited. This due, part, current spread, unintended axonal activation, and limited resolution existing devices. Here we show, using recent model prosthetic vision, optimizing how stimuli encoded device help overcome some these limitations, leading dramatic improvements perception. Approach We propose do this practice, patients’ feedback task. The main challenge our approach comes from fact that, typically, one only has access number noisy responses patients. two ways deal this: first, use vision constrain simplify optimization. show if knew parameters given patient, it would possible greatly improve their perceptual performance. Second preferential Bayesian optimization efficiently learn each minimal trials. Main results To test approach, presented healthy subjects generated replicate experience fitted an implant. Our procedure led significant robust perceived image quality, transferred increased performance other tasks. Significance Importantly, agnostic type prosthesis thus could readily implemented implants.

Язык: Английский

Bayesian Optimization in Drug Discovery DOI
Lionel Colliandre, Christophe Müller

Methods in molecular biology, Год журнала: 2023, Номер unknown, С. 101 - 136

Опубликована: Сен. 13, 2023

Язык: Английский

Процитировано

8
Optimized Ensemble Support Vector Regression Models for Predicting Stock Prices with Multiple Kernels DOI Creative Commons
Subba Reddy Thumu, N. Geethanjali

Acta Informatica Pragensia, Год журнала: 2024, Номер 13(1), С. 24 - 37

Опубликована: Янв. 14, 2024

Stock forecasting is a complicated and daily challenge for investors because of the non-linearity market high volatility financial assets such as stocks, bonds other commodities.There need powerful adaptive stock prediction model that handles complexities provides accurate predictions.The support vector regression (SVR) one most prominent machine learning models time series data.An ensemble hyperbolic tangent kernel SVR (HTK-SVR-BO) proposed in this paper, combining Tanh inverse kernels with Bayesian optimization.Combining strengths multiple using technique then optimization to identify optimal values each enhance performance possible.Our compared an (LPR-SVR-BO), which uses well-known types, including linear, polynomial radial basis function (RBF).We apply Microsoft Corporation (MSFT) prices.The mean absolute error (MAE), squared (MSE), root (RMSE), R2 score (model accuracy) percentage (MAPE) are metrics used compare effectiveness model.In our comparison, HTK-SVR-BO performs better terms LPR-SVR-BO achieves results 0.27424, 0.13392, 0.36595, 0.99997 5.2331 respectively.According analysis, more predictive may generalize previously unknown data effectively, so it can be when future prices.

Язык: Английский

Процитировано

3
基于贝叶斯优化XGBoost 模型的黏土压缩指标预测 DOI
Hongtao Wu, Z. L. Zhang, Daniel Dias

и другие.

Journal of Central South University, Год журнала: 2024, Номер 31(11), С. 3914 - 3929

Опубликована: Июнь 20, 2024

Язык: Английский

Процитировано

3
Transferable and data efficient metamodeling of storm water system nodal depths using auto-regressive graph neural networks DOI Creative Commons
Alexander Garzón, Zoran Kapelan, Jeroen Langeveld

и другие.

Water Research, Год журнала: 2024, Номер 266, С. 122396 - 122396

Опубликована: Сен. 11, 2024

Язык: Английский

Процитировано

3
Human-in-the-loop optimization of visual prosthetic stimulation DOI
Tristan Fauvel, Matthew Chalk

Journal of Neural Engineering, Год журнала: 2022, Номер 19(3), С. 036038 - 036038

Опубликована: Июнь 1, 2022

Abstract Objective . Retinal prostheses are a promising strategy to restore sight patients with retinal degenerative diseases. These devices compensate for the loss of photoreceptors by electrically stimulating neurons in retina. Currently, visual function that can be recovered such is very limited. This due, part, current spread, unintended axonal activation, and limited resolution existing devices. Here we show, using recent model prosthetic vision, optimizing how stimuli encoded device help overcome some these limitations, leading dramatic improvements perception. Approach We propose do this practice, patients’ feedback task. The main challenge our approach comes from fact that, typically, one only has access number noisy responses patients. two ways deal this: first, use vision constrain simplify optimization. show if knew parameters given patient, it would possible greatly improve their perceptual performance. Second preferential Bayesian optimization efficiently learn each minimal trials. Main results To test approach, presented healthy subjects generated replicate experience fitted an implant. Our procedure led significant robust perceived image quality, transferred increased performance other tasks. Significance Importantly, agnostic type prosthesis thus could readily implemented implants.

Язык: Английский

Процитировано

12