Automatic Induction of Neural Network Decision Tree Algorithms

Siu, Chapman

doi:10.1007/978-3-030-22871-2_48

Cited by 3 publications

(4 citation statements)

References 4 publications

(8 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Learning Decision Stumps as a Model Selection Problem If we interpret the decision function to be a model selection process, then model selection approaches can be used to determine the ideal model, and hence axis-parallel split for the decision function. One approach is to use a stacking model, which has been shown to asymptotically approach Bayesian model averaging and can also be used as a neural network architecture search problem for neural network decision tree algorithms [16]. We then formulate stacking every potential axis-parallel split as follows:…”

Section: Decision Stumpsmentioning

confidence: 99%

“…TreeGrad is an extension of Deep Neural Decision Forests, which treats the node split structure to be a neural network architecture search problem in the manner described by [16]; whilst enforcing neural network compression approaches to render our decision trees to be more interpretable through creating axis-parallel splits.…”

Section: Related Workmentioning

confidence: 99%

“…Stacking models are an ensemble model in the form ŷ = k v k M k , for set of real weights v k , for all k representing the universe of candidate models [18] [2]. It has been shown to asymptotically approach Bayesian model averaging and can also be used as a neural architecture search problem for neural network decision tree algorithms [16].…”

Section: Ensemble Learningmentioning

confidence: 99%

See 2 more Smart Citations

Transferring Tree Ensembles to Neural Networks

Siu¹

2019

Neural Information Processing

Self Cite

View full text Add to dashboard Cite

Gradient Boosting Decision Tree (GBDT) is a popular machine learning algorithms with implementations such as LightGBM and in popular machine learning toolkits like Scikit-Learn. Many implementations can only produce trees in an offline manner and in a greedy manner. We explore ways to convert existing GBDT implementations to known neural network architectures with minimal performance loss in order to allow decision splits to be updated in an online manner and provide extensions to allow splits points to be altered as a neural architecture search problem. We provide learning bounds for our neural network and demonstrate that our non-greedy approach has comparable performance to state-of-the-art offline, greedy tree boosting models.

show abstract

Section: Decision Stumpsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Ensemble Learningmentioning

confidence: 99%

See 1 more Smart Citation

Transferring Tree Ensembles to Neural Networks

Siu¹

2019

Neural Information Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…In this study, we conducted a comprehensive investigation on how ML algorithms can be exploited to cope with the challenges addressed above. We analyzed how different types of prediction models, namely the conventional decision tree (DT) 21,22 prediction models, the state-of-the-art Deep Neural Network (DNN) prediction models 15 , as well as the logistic regression (LR) based prediction models, performed in identifying the patients with high risk of rapid progression and death. With the experimental results, our discussions have focused on how DT prediction models can be incorporated to facilitate physicians' decisions on triaging patients based on risk of clinical deterioration and prioritizing scarce medical resources for best population outcomes during a pandemic of EID.…”

Section: Introductionmentioning

confidence: 99%

Supervised Machine Learning for Risk Stratification of Influenza-Like Illness: A Model to Prioritize Emerging Infectious Disease Disaster Responses

Chiu

Hwang

Chen

et al. 2021

Preprint

View full text Add to dashboard Cite

Emerging infectious diseases (EIDs), including the latest COVID-19 pandemic, have emerged and raised global public health crises in recent decades. Without existing protective immunity, an EID may spread rapidly and cause mass casualties in a very short time. Therefore, it is imperative to identify cases with risk of disease progression for the best allocation of medical resources in case medical facilities are overwhelmed with a flood of patients. This study aimed to exploit machine learning technologies to cope with this challenge. The study was based on 83,227 hospital admissions with influenza-like illness and we analysed the risk effects of 19 comorbidities along with age and gender for severe illness or mortality risk. The experimental results revealed that the conventional decision tree (DT) models built with only 6 features, including age, gender, and four comorbidities, delivered the same level of prediction accuracy as the state-of-the-art deep neural network models built with 18 features. Accordingly, we further studied how to exploit the DT models with different sensitivity levels to determine patient triage and optimize medical resource allocation in different stages of an EID disaster to aid the frontline clinicians and policy-makers. In conclusion, our study demonstrated an approach to exploit machine learning technologies to cope with the challenges during the outbreak of an EID.

show abstract

Machine learning for emerging infectious disease field responses

Chiu

Hwang

Chen

et al. 2022

Sci Rep

View full text Add to dashboard Cite

Emerging infectious diseases (EIDs), including the latest COVID-19 pandemic, have emerged and raised global public health crises in recent decades. Without existing protective immunity, an EID may spread rapidly and cause mass casualties in a very short time. Therefore, it is imperative to identify cases with risk of disease progression for the optimized allocation of medical resources in case medical facilities are overwhelmed with a flood of patients. This study has aimed to cope with this challenge from the aspect of preventive medicine by exploiting machine learning technologies. The study has been based on 83,227 hospital admissions with influenza-like illness and we analysed the risk effects of 19 comorbidities along with age and gender for severe illness or mortality risk. The experimental results revealed that the decision rules derived from the machine learning based prediction models can provide valuable guidelines for the healthcare policy makers to develop an effective vaccination strategy. Furthermore, in case the healthcare facilities are overwhelmed by patients with EID, which frequently occurred in the recent COVID-19 pandemic, the frontline physicians can incorporate the proposed prediction models to triage patients suffering minor symptoms without laboratory tests, which may become scarce during an EID disaster. In conclusion, our study has demonstrated an effective approach to exploit machine learning technologies to cope with the challenges faced during the outbreak of an EID.

show abstract

Automatic Induction of Neural Network Decision Tree Algorithms

Cited by 3 publications

References 4 publications

Transferring Tree Ensembles to Neural Networks

Transferring Tree Ensembles to Neural Networks

Supervised Machine Learning for Risk Stratification of Influenza-Like Illness: A Model to Prioritize Emerging Infectious Disease Disaster Responses

Machine learning for emerging infectious disease field responses

Contact Info

Product

Resources

About