Create ethical and explainable machine learning algorithms through the closest thing machine learning outputs have to a p-value right now.
This repository expands on Google/TrustScore: To trust or not to trust a classifier. A measure of uncertainty for any trained (possibly Black-box) classifier which is more effective than the classifier's own implied confidence (e.g. softmax probability for a neural network) by adding comments to their base python code, trustscores.py, and providing the necessary code to find trust scores in R using a package that converts the existing Python code to R.
Much of this repository is based on To Trust Or Not To Trust A Classifier's concepts and code. All references are cited below.
Jiang, H., & Hembise, C. (n.d.). Google/TrustScore: To trust or not to trust a classifier. A measure of uncertainty for any trained (possibly Black-box) classifier which is more effective than the classifier's own implied confidence (e.g. softmax probability for a neural network). GitHub. Retrieved December 22, 2021, from https://github.com/google/TrustScore
Jiang, H., Kim, B., Guan, M. Y., & Gupta, M. (2018, October 26). To trust or not to trust a classifier. arXiv.org. Retrieved December 22, 2021, from https://arxiv.org/abs/1805.11783
Gajane, P., & Pechenizkiy, M. (2018, May 28). On formalizing fairness in prediction with machine learning. arXiv.org. Retrieved December 22, 2021, from https://arxiv.org/abs/1710.03184
Papernot, N., & McDaniel, P. (2018, March 13). Deep K-nearest neighbors: Towards confident, interpretable and robust deep learning. arXiv.org. Retrieved December 22, 2021, from https://arxiv.org/abs/1803.04765v1
Interface to python. Interface to Python • reticulate. (n.d.). Retrieved December 22, 2021, from https://rstudio.github.io/reticulate/
Yahya, & Mortensen, P. (2021, May 8). Convergencewarning: Lbfgs failed to converge (status=1): Stop: Total no. of iterations reached limit. Stack Overflow. Retrieved December 22, 2021, from https://stackoverflow.com/a/62659927
This teachable unit is created from the references listed above, as well as my own work. It is offered under the Apache Licence.
Copyright 2021 Sarah Gillespie
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
THe original Python code about trust scores is sourced from https://github.com/google/TrustScore, also with an Apache License.