GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → williamcaicedo → ISeeU

williamcaicedo / ISeeU

Licence: MIT, MIT licenses found Licenses found MIT LICENSE MIT license.txt
ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU

Programming Languages

Jupyter Notebook
11667 projects

Labels

deep-learningconvolutional-neural-networksintensive-careinterpretable-deep-learning

Projects that are alternatives of or similar to ISeeU

deep-explanation-penalization
Code for using CDEP from the paper "Interpretations are useful: penalizing explanations to align neural networks with prior knowledge" https://arxiv.org/abs/1909.13584
Stars: ✭ 110 (+450%)
Mutual labels:  interpretable-deep-learning
m-phate
Multislice PHATE for tensor embeddings
Stars: ✭ 54 (+170%)
Mutual labels:  interpretable-deep-learning
redunet paper
Official NumPy Implementation of Deep Networks from the Principle of Rate Reduction (2021)
Stars: ✭ 49 (+145%)
Mutual labels:  interpretable-deep-learning
self critical vqa
Code for NeurIPS 2019 paper ``Self-Critical Reasoning for Robust Visual Question Answering''
Stars: ✭ 39 (+95%)
Mutual labels:  interpretable-deep-learning
Awesome Machine Learning Interpretability
A curated list of awesome machine learning interpretability resources.
Stars: ✭ 2,404 (+11920%)
Mutual labels:  interpretable-deep-learning
Pytorch Grad Cam
Many Class Activation Map methods implemented in Pytorch for CNNs and Vision Transformers. Including Grad-CAM, Grad-CAM++, Score-CAM, Ablation-CAM and XGrad-CAM
Stars: ✭ 3,814 (+18970%)
Mutual labels:  interpretable-deep-learning
Layerwise-Relevance-Propagation
Implementation of Layerwise Relevance Propagation for heatmapping "deep" layers
Stars: ✭ 78 (+290%)
Mutual labels:  interpretable-deep-learning
CNN-Units-in-NLP
✂️ Repository for our ICLR 2019 paper: Discovery of Natural Language Concepts in Individual Units of CNNs
Stars: ✭ 26 (+30%)
Mutual labels:  interpretable-deep-learning
InterpretableCNN
No description or website provided.
Stars: ✭ 36 (+80%)
Mutual labels:  interpretable-deep-learning
path explain
A repository for explaining feature attributions and feature interactions in deep neural networks.
Stars: ✭ 151 (+655%)
Mutual labels:  interpretable-deep-learning
ProtoTree
ProtoTrees: Neural Prototype Trees for Interpretable Fine-grained Image Recognition, published at CVPR2021
Stars: ✭ 47 (+135%)
Mutual labels:  interpretable-deep-learning
ShapleyExplanationNetworks
Implementation of the paper "Shapley Explanation Networks"
Stars: ✭ 62 (+210%)
Mutual labels:  interpretable-deep-learning
rl singing voice
Unsupervised Representation Learning for Singing Voice Separation
Stars: ✭ 18 (-10%)
Mutual labels:  interpretable-deep-learning

ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU

Note: This version of ISeeU has been tested with numpy 1.12.1, pandas 0.23.4, keras 2.2.4, deeplift 0.6.6.2, matplotlib 2.0.2 and tensorflow 1.9.0.

A ConvNet trained on MIMIC-III data for mortality prediction inside the Intensive Care Unit. It uses a set of 22 predictors sampled during the first 48h of ICU stay to predict the probability of mortality. This set of predictors roughly corresponds to those used by the SAPS-II severity score:

  • AGE
  • AIDS
  • BICARBONATE
  • BILIRRUBIN
  • BUN
  • DIASTOLIC BP
  • ELECTIVE
  • FiO2
  • GCSEyes
  • GCSMotor
  • GCSVerbal
  • HEART RATE
  • LYMPHOMA
  • METASTATIC CANCER
  • PO2
  • POTASSIUM
  • SODIUM
  • SURGICAL
  • SYSTOLIC BP
  • TEMPERATURE
  • URINE OUTPUT
  • WBC

ISeeU achieves 0.8735 AUROC when evaluated on MIMIC-III. More information is available in our paper. It also can be installed from PyPi:

pip install iseeu
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].