Explaining the Predictions of Unsupervised Learning Models

Montavon, Grégoire; Kauffmann, Jacob R.; Samek, Wojciech; Mueller, K.

doi:10.1007/978-3-031-04083-2_7

Cited by 16 publications

(5 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The primary criterion for choosing an ML method in the current research was explainability-an explicit relationship between input features and an output data model [53]. Unsupervised ML algorithms based on using distance metrics between data points in hyperspace usually provide unsatisfactory results, while the results of applying hierarchical, density-based techniques are hard to explain [54]. A general rule for ML methods is that the higher complexity of the model, the more difficult interpretation is.…”

Section: Nonparametric Methods O(2n 3 )mentioning

confidence: 99%

Terahertz Time-Domain Spectroscopy of Blood Serum for Differentiation of Glioblastoma and Traumatic Brain Injury

Vrazhnov,

Ovchinnikova,

Kabanova

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

The possibility of the differentiation of glioblastoma from traumatic brain injury through blood serum analysis by terahertz time-domain spectroscopy and machine learning was studied using a small animal model. Samples of a culture medium and a U87 human glioblastoma cell suspension in the culture medium were injected into the subcortical brain structures of groups of mice referred to as the culture medium injection groups and glioblastoma groups, accordingly. Blood serum samples were collected in the first, second, and third weeks after the injection, and their terahertz transmission spectra were measured. The injection caused acute inflammation in the brain during the first week, so the culture medium injection group in the first week of the experiment corresponded to a traumatic brain injury state. In the third week of the experiment, acute inflammation practically disappeared in the culture medium injection groups. At the same time, the glioblastoma group subjected to a U87 human glioblastoma cell injection had the largest tumor size. The THz spectra were analyzed using two dimensionality reduction algorithms (principal component analysis and t-distributed Stochastic Neighbor Embedding) and three classification algorithms (Support Vector Machine, Random Forest, and Extreme Gradient Boosting Machine). Constructed prediction data models were verified using 10-fold cross-validation, the receiver operational characteristic curve, and a corresponding area under the curve analysis. The proposed machine learning pipeline allowed for distinguishing the traumatic brain injury group from the glioblastoma group with 95% sensitivity, 100% specificity, and 97% accuracy with the Extreme Gradient Boosting Machine. The most informative features for these groups’ differentiation were 0.37, 0.40, 0.55, 0.60, 0.70, and 0.90 THz. Thus, an analysis of mouse blood serum using terahertz time-domain spectroscopy and machine learning makes it possible to differentiate glioblastoma from traumatic brain injury.

show abstract

Section: Nonparametric Methods O(2n 3 )mentioning

confidence: 99%

Terahertz Time-Domain Spectroscopy of Blood Serum for Differentiation of Glioblastoma and Traumatic Brain Injury

Vrazhnov,

Ovchinnikova,

Kabanova

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…Supervised learning utilizes labeled data to train a model for making predictions or classifications [8,9]. Unsupervised learning explores unlabeled data to discover hidden structures and relationships [10]. Reinforcement learning focuses on an agent learning through interactions with an environment to maximize cumulative rewards and make sequential decisions [11].…”

Section: Machine Learningmentioning

confidence: 99%

Developing an Interferogram-Based Module with Machine Learning for Maintaining Leveling of Glass Substrates

Chang

Wang

et al. 2023

Machines

View full text Add to dashboard Cite

In this research, we propose a method that utilizes machine learning to maintain the parallelism of the resonant cavity in a Fabry–Perot interferometer designed specifically for glass substrates. Based on the optical principle and theory, we establish a proportional relationship between interference fringes and the inclination angle of the mirrors. This enables an accurate determination of the inclination angle using supervised learning, specifically classification. By training a machine learning model with labeled data, interference fringe patterns are categorized into three levels, with approximately 100 training data available for each level in each location. The experimental results of Level 2 and Level 3 classification indicate an average number of corrections of 2.55 and 3.55 times, respectively, in achieving the target position with a correction error of less than 30 arc seconds. These findings demonstrate the essential nature of this parallelism maintenance technology for the semiconductor industry and precision mechanical engineering.

show abstract

“…Supervised learning involves training a model on labeled data, where the correct output is provided for each input. Unsupervised learning involves training a model on unlabeled data, where the goal is to discover hidden patterns or structures in the data [9]. Reinforcement learning involves training a model to make decisions in an environment by receiving feedback in the form of rewards or punishments [10].…”

Section: Machine Learningmentioning

confidence: 99%

Leveling Maintenance Mechanism by Using the Fabry-Perot Interferometer with Machine Learning Technology

Chang

Wang

et al. 2023

Teh. glas. (Online)

View full text Add to dashboard Cite

This study proposes a method for maintaining parallelism of the optical cavity of a laser interferometer using machine learning. The Fabry-Perot interferometer is utilized as an experimental optical structure in this research due to its advantage of having a brief optical structure. The supervised machine learning method is used to train algorithms to accurately classify and predict the tilt angle of the plane mirror using labeled interference images. Based on the predicted results, stepper motors are fixed on a plane mirror that can automatically adjust the pitch and yaw angles. According to the experimental results, the average correction error and standard deviation in 17-grid classification experiment are 32.38 and 11.21 arcseconds, respectively. In 25-grid classification experiment, the average correction error and standard deviation are 19.44 and 7.86 arcseconds, respectively. The results show that this parallelism maintenance technology has essential for the semiconductor industry and precision positioning technology.

show abstract

Explaining the Predictions of Unsupervised Learning Models

Cited by 16 publications

References 31 publications

Terahertz Time-Domain Spectroscopy of Blood Serum for Differentiation of Glioblastoma and Traumatic Brain Injury

Terahertz Time-Domain Spectroscopy of Blood Serum for Differentiation of Glioblastoma and Traumatic Brain Injury

Developing an Interferogram-Based Module with Machine Learning for Maintaining Leveling of Glass Substrates

Leveling Maintenance Mechanism by Using the Fabry-Perot Interferometer with Machine Learning Technology

Contact Info

Product

Resources

About