Filters








12,140 Hits in 3.0 sec

On Explaining Random Forests with SAT [article]

Yacine Izza, Joao Marques-Silva
2021 arXiv   pre-print
Random Forest (RFs) are among the most widely used Machine Learning (ML) classifiers.  ...  This contrasts with earlier work on explaining boosted trees (BTs) and neural networks (NNs), which requires encodings based on SMT/MILP.  ...  Conclusion This paper proposes a novel approach for explaining random forests.  ... 
arXiv:2105.10278v1 fatcat:w5xkh63q6bbmnnpkufrntysxbq

On Explaining Random Forests with SAT

Yacine Izza, Joao Marques-Silva
2021 Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence   unpublished
Random Forest (RFs) are among the most widely used Machine Learning (ML) classifiers.  ...  This contrasts with earlier work on explaining boosted trees (BTs) and neural networks (NNs), which requires encodings based on SMT/MILP.  ...  Conclusion This paper proposes a novel approach for explaining random forests.  ... 
doi:10.24963/ijcai.2021/356 fatcat:lvhphglnyzdldoy42p7swb4r4a

ASTERYX : A model-Agnostic SaT-basEd appRoach for sYmbolic and score-based eXplanations [article]

Ryma Boumazouza
2022 arXiv   pre-print
Our approach is declarative and it is based on the encoding of the model to be explained in an equivalent symbolic representation, this latter serves to generate in particular two types of symbolic explanations  ...  This paper proposes a generic agnostic approach named ASTERYX allowing to generate both symbolic explanations and score-based ones.  ...  Explaining random forests and decision trees is dealt with for instance in [2] and [15, 18] respectively.  ... 
arXiv:2206.11900v1 fatcat:4noywv55g5efxm5ss2vdtcso2u

A Model-Agnostic SAT-based Approach for Symbolic Explanation Enumeration [article]

Ryma Boumazouza
2022 arXiv   pre-print
More precisely, we generate explanations to locally explain a single prediction by analyzing the relationship between the features and the output.  ...  The experimental results on image classification task show the feasibility of the proposed approach and its effectiveness in providing Sufficient Reasons and Counterfactuals explanations.  ...  CNF encoding of random forests In this work, we adopted the random forest 3 as the surrogate model f S .  ... 
arXiv:2206.11539v1 fatcat:tg4fenvtlnem7hlfztcllj7iqa

Understanding the empirical hardness ofNP-complete problems

Kevin Leyton-Brown, Holger H. Hoos, Frank Hutter, Lin Xu
2014 Communications of the ACM  
One of the most popular approaches for the formal verification of hardware and software relies on generalpurpose SAT solvers and SAT encodings, typically with hundreds of thousands of variables.  ...  The focus of this paper is on ways that EHMs contribute to our understanding of NP-complete problems; however, they are also 1 Some work described in this article was performed with additional coauthors  ...  The dotted line shows p(c, v) for uniform-random 3-SAT instances with v = 400, while the solid line shows the mean runtime of march_hi [11] , one of the best SAT solvers for uniform-random 3-SAT, on the  ... 
doi:10.1145/2594413.2594424 fatcat:t2chnugwqrg3rcsa64yynxkftq

Combining randomized field experiments with observational satellite data to assess the benefits of crop rotations on yields [article]

Dan M. Kluger, Art B. Owen, David B. Lobell
2021 arXiv   pre-print
Further, the causal estimates based on our method suggest that benefits of crop rotations on corn yield are lower in years and locations with high temperatures whereas the benefits of crop rotations on  ...  soy yield are higher in years and locations with high temperatures.  ...  Finally, the authors thank Jill Deines for assistance with extracting the observational data from Google Earth Engine.  ... 
arXiv:2112.13700v1 fatcat:xqhr2bu2ujfolhpuuq3242czuq

Exploiting Polarity Features for Developing Sentiment Analysis Tool

Lubna Zafar, Muhammad Tanvir Afzal, Usman Ahmed
2017 Extended Semantic Web Conference  
It is found that adverb, adjectives, and verb combination can achieve the nest accuracy when trained on a specific settings of Random Forest Classifier and Gradient Boosting Classifier.  ...  This paper explains the lessons learned from the literature and followed by the findings and it gives an input to build a scalable system:  ...  Gradient Boosting and Random Forest classifiers gave 0.81 precision on Adjective-Adverb and Verb Combination.  ... 
dblp:conf/esws/ZafarAA17 fatcat:roo2fdgr2vhu5fkhyhteb7vybm

Influence of Environmental Factors on Forest Understorey Species in Northern Mexico

Juan F. Maciel-Nájera, M. Socorro González-Elizondo, José Ciro Hernández-Díaz, Carlos A. López-Sánchez, Claudia Edith Bailón-Soto, Artemio Carrillo-Parra, Christian Wehenkel
2021 Forests  
Data were analyzed using a Binomial Logistic Model (BLM) and Random Forest (RF) classification.  ...  forests of northern Mexico.  ...  Santos Gregorio Rodríguez García, of the ejido El Largo y Anexos and the Regional Forest Management Unit (UMAFOR 0802) from Madera, Chihuahua for the information provided. We also thank Ing.  ... 
doi:10.3390/f12091198 fatcat:zdimiq7omjewdle437doag7m7i

A machine learning approach for MSG/SEVIRI SST bias estimation

Stéphane Saux-Picart
2017 Zenodo  
Objective Design statistical models to represent ∆SST = SST sat − SST buoys with a set of explaining variables. We consider in-situ measurements from drifting buoys to be the ground truth.  ...  SST is the model estimate of ∆SST, ∆SST is the average of ∆SST, n is the number of observation and p is the number of explaining variables. An example using random forest model: 15/11/2014 00h.  ... 
doi:10.5281/zenodo.5163853 fatcat:bk3pu6g6czacvhnvs23axyxo24

Validation of Visually Interpreted Corine Land Cover Classes with Spectral Values of Satellite Images and Machine Learning

Orsolya Gyöngyi Varga, Zoltán Kovács, László Bekő, Péter Burai, Zsuzsanna Csatáriné Szabó, Imre Holb, Sarawut Ninsawat, Szilárd Szabó
2021 Remote Sensing  
We then performed Linear Discriminant Analysis (LDA) and Random Forest classifications to reveal if CLC L1 level categories were confirmed by spectral values.  ...  We proved the representativeness of the results with a repeated randomized test, and only PlanetScope seemed to be ungeneralizable.  ...  ) of 10 randomized Random Forest classifications.  ... 
doi:10.3390/rs13050857 fatcat:n2tg55dad5dlxmchyo5mcsiu6m

Exploring Machine Learning to Correct Satellite-Derived Sea Surface Temperatures

Stéphane Saux Picart, Pierre Tandeo, Emmanuelle Autret, Blandine Gausset
2018 Remote Sensing  
Four regression models are used: Simple multi-linear regression, Least Square Shrinkage and Selection Operator (LASSO), Generalised Additive Model (GAM) and random forest.  ...  In the case of geostationary satellites for which a large number of collocations is available, results show that the random forest model is the best model to predict the systematic errors and it is computationally  ...  Acknowledgments: The data from the EUMETSAT Satellite Application Facility on Ocean and Sea Ice used in this study are accessible through the SAF's homepage http://osi-saf.eumetsat.int.  ... 
doi:10.3390/rs10020224 fatcat:sfud3d23xbcsdcn5zh66dqbbue

Efficiently Calculating Evolutionary Tree Measures Using SAT [chapter]

María Luisa Bonet, Katherine St. John
2009 Lecture Notes in Computer Science  
We focus on two popular comparison measures for trees: the hybridization number and the rooted subtree-prune-and-regraft (rSPR) distance.  ...  We use state-of-the-art SAT solvers to determine if the formula encoding the measure has a satisfying assignment.  ...  Notice that we compare all kinds of different solvers: local search algorithms (the first three), DPLL with learning (minisat), SAT solver portfolio (SATzilla) and solver specialized on random instances  ... 
doi:10.1007/978-3-642-02777-2_3 fatcat:e7vxsy3yonhxrktnknjzrfi5fy

Stand age and species richness dampen interannual variation of ecosystem-level photosynthetic capacity

Talie Musavi, Mirco Migliavacca, Markus Reichstein, Jens Kattge, Christian Wirth, T. Andrew Black, Ivan Janssens, Alexander Knohl, Denis Loustau, Olivier Roupsard, Andrej Varlagin, Serge Rambal (+5 others)
2017 Nature Ecology & Evolution  
We find that the IAV of GPP sat is greatly reduced in older and more diverse forests, and is higher in younger forests with few dominant species.  ...  Older and more diverse forests seem to dampen the effect of climate variability on the carbon cycle irrespective of forest type.  ...  Stand age, which is negatively correlated with cvGPP sat , is the most important predictor (from the 55% explained variance by both variables, the relative contribution to the explained variance by stand  ... 
doi:10.1038/s41559-016-0048 pmid:28812604 fatcat:b4wuaf3f4rdpve25xfb5oupqhy

DeepSat

Saikat Basu, Sangram Ganguly, Supratik Mukhopadhyay, Robert DiBiano, Manohar Karki, Ramakrishna Nemani
2015 Proceedings of the 23rd SIGSPATIAL International Conference on Advances in Geographic Information Systems - GIS '15  
Comparative studies with a Random Forest classifier show the advantage of an unsupervised learning approach over traditional supervised learning techniques.  ...  On SAT-6, it produces a classification accuracy of 93.9% and outperforms the other algorithms by ∼15%.  ...  On SAT-4, the Random forest classifier produces an accuracy of 69% while on SAT-6, it produces an accuracy of 54%. The highest accuracy was obtained for a forest with 100 trees.  ... 
doi:10.1145/2820783.2820816 dblp:conf/gis/BasuGMDKN15 fatcat:zzkxeczv4vgz3exd6h5da46z3q

DeepSat - A Learning framework for Satellite Imagery [article]

Saikat Basu, Sangram Ganguly, Supratik Mukhopadhyay, Robert DiBiano, Manohar Karki, Ramakrishna Nemani
2015 arXiv   pre-print
Comparative studies with a Random Forest classifier show the advantage of an unsupervised learning approach over traditional supervised learning techniques.  ...  On SAT-6, it produces a classification accuracy of 93.9% and outperforms the other algorithms by ~15%.  ...  On SAT-4, the Random forest classifier produces an accuracy of 69% while on SAT-6, it produces an accuracy of 54%. The highest accuracy was obtained for a forest with 100 trees.  ... 
arXiv:1509.03602v1 fatcat:a7hprfmomba6hn5bnwg2cvdu6a
« Previous Showing results 1 — 15 out of 12,140 results