An Explainable Bayesian Decision Tree Algorithm

Nuti, Giuseppe; Rugama, Lluís Antoni Jiménez; Cross, Andreea-Ingrid

doi:10.3389/fams.2021.598833

Cited by 10 publications

(10 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our previous study we used the Joint and Individual Variance Explained (JIVE) approach to categorically assign transcriptomes to sex-specific subtypes of glioblastoma [22]. An advantage of the Bayesian Nearest Neighbor (BNN) algorithm is that it can infer "breakpoints" between local groupings of nearest neighbors and estimate individual TSI values for any transcriptome along a continuous spectrum of values as a Bayesian posterior probability using that transcriptomes' local neighbors [24].…”

Section: Inferring Transcriptomic Sex Index (Tsi) Using Bayesian Near...mentioning

confidence: 99%

Accounting for sex differences variability in the design of sex-adapted cancer treatments

Yang

Rubin

2023

Preprint

View full text Add to dashboard Cite

The significant sex differences that exist in cancer mechanisms, incidence, and survival, have yet to impact clinical practice. We hypothesized that one barrier to translation is that sex differences in cancer phenotypes resemble sex differences in height: highly overlapping, but distinct, male and female population distributions that vary continuously between female- and male- biased extremes. A consequence of this variance is that sex-specific treatments are rendered unrealistic, and our translational goal should be adaptation of treatment to the unique mix of sex-biased mechanisms that are present in each patient. To develop a tool that could advance this goal, we applied a Bayesian Nearest Neighbor (BNN) analysis to 8370 cancer transcriptomes from 26 different adult and 4 different pediatric cancer types to establish patient-specific Transcriptomic Sex Indices (TSI). TSI precisely partitions an individual patients whole transcriptome into female- and male- biased components such that cancer type, patient sex, and transcriptomics, provide a novel and patient-specific mechanistic identifier that can be used for sex-adapted, precision cancer treatment planning.

show abstract

Section: Inferring Transcriptomic Sex Index (Tsi) Using Bayesian Near...mentioning

confidence: 99%

Accounting for sex differences variability in the design of sex-adapted cancer treatments

Yang

Rubin

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Each transition between nodes depends on a condition. The condition mentioned here is the theory on which the chosen algorithm is based [21][22]. Decision trees are very advantageous for reasons such as low calculation cost and ease of understanding.…”

Section: Random Tree Algorithmmentioning

confidence: 99%

Analysis and detection of Titanic survivors using generalized linear models and decision tree algorithm

Durmuş

Güneri

2020

International Journal of Applied Mathematics Electronics and Computers

View full text Add to dashboard Cite

In the article, it is aimed to investigate the factors affecting survival in today's legendary giant accident with different methods. The analysis aims to find the method that best determines survival. For this purpose, logit and probit models from generalized linear models and random tree algorithm from decision tree methods were used. The study was carried out in two stages. Firstly; in the analysis made with generalized linear models, variables that did not contribute significantly to the model were determined. Classification accuracy was found to be 79.89% for the logit model and 79.04% for the probit model. In the second stage; classification analysis was performed with random tree decision trees. Classification accuracy was determined to be 77.21%. In addition; according to the results obtained from the generalized linear models, the classification analysis was repeated by removing the data that made meaningless contribution to the model. The classification rate increased by 4.36% and reached 81.57%. After all; It was determined that the decision tree analysis made with the variables extracted from the model gave better results than the analysis made with the original variables. These results are thought to be useful for researchers working on classification analysis. In addition, the results can be used for purposes such as data preprocessing, data cleaning.

show abstract

“…Our overall objective is to estimate the probability of an outcome y ∈ R given the input x ∈ R d , namely p(y|x). Our approach is based on a soft version of the Bayesian Decision Trees from [14] which leverages the optimization approach of neural networks.…”

Section: Adaptive Bayesian Reticulummentioning

confidence: 99%

“…In other words, we only consider affine hyperplanes and choose s to be the sigmoid function. The Bayesian Reticulum is a Reticulum where we define the expected log-likelihood and p(y|x ∈ ) for each leaf given the input data, as introduced in [14].…”

Section: Bayesian Reticulummentioning

confidence: 99%

See 1 more Smart Citation

Adaptive Bayesian Reticulum

Nuti¹,

Rugama²,

Thommen³

2019

Preprint

Self Cite

View full text Add to dashboard Cite

Neural Networks and Decision Trees: two popular techniques for supervised learning that are seemingly disconnected in their formulation and optimization method, have recently been combined in a single construct. The connection pivots on assembling an artificial Neural Network with nodes that allow for a gate-like function to mimic a tree split, optimized using the standard approach of recursively applying the chain rule to update its parameters. Yet two main challenges have impeded wide use of this hybrid approach: (a) the inability of global gradient ascent techniques to optimize hierarchical parameters (as introduced by the gate function); and (b) the construction of the tree structure, which has relied on standard decision tree algorithms to learn the network topology or incrementally (and heuristically) searching the space at random. Here we propose a probabilistic construct that exploits the idea of a node's unexplained potential (the total error channeled through the node) in order to decide where to expand further, mimicking the standard tree construction in a Neural Network setting, alongside a modified gradient ascent that first locally optimizes an expanded node before a global optimization. The probabilistic approach allows us to evaluate each new split as a ratio of likelihoods that balances the statistical improvement in explaining the evidence against the additional model complexity -thus providing a natural stopping condition. The result is a novel classification and regression technique that leverages the strength of both: a tree-structure that grows naturally and is simple to interpret with the plasticity of Neural Networks that allow for soft margins and slanted boundaries.

show abstract

An Explainable Bayesian Decision Tree Algorithm

Cited by 10 publications

References 15 publications

Accounting for sex differences variability in the design of sex-adapted cancer treatments

Accounting for sex differences variability in the design of sex-adapted cancer treatments

Analysis and detection of Titanic survivors using generalized linear models and decision tree algorithm

Adaptive Bayesian Reticulum

Contact Info

Product

Resources

About