数据科学领域常用技术

统计于最新kaggle-survey-2021

• 工作内容

1.图片分类、其他网络 VGG, Inception, ResNet, ResNeXt, NASNet, EfficientNet CNN) 2.目标检测(YOLOv3, RetinaNet, etc)

3.常用图片视频工具 (PIL, cv2, skimage, etc)

4.生成对抗网络 (GAN, VAE, etc)

• 学习途径

1. Coursera、Kaggle

• 语言

1. Python、SQL

• 可视化工具

1.Tableau

• 计算引擎

1. Google Cloud Compute Engine

2. Microsoft Azure Virtual Machines

• 服务器

1. AWS、Google Cloud Platform (GCP)

• 框架

1. TensorFlow、Keras、PyTorch、Xgboost、LightGBM、CatBoost、PyTorch Lightning

• 文本工具

1. Word embeddings/vectors (GLoVe, fastText, word2vec)

• 自动化工具

1. Automation of full ML pipelines (e.g. Google Cloud AutoML, H2O Driverless AI)

2. Automated feature engineering/selection (e.g. tpot, boruta_py)

3. Automated data augmentation (e.g. imgaug, albumentations)

4. Automated hyperparameter tuning (e.g. hyperopt, ray.tune, Vizier)

5. Automated model architecture searches (e.g. darts, enas)

• 开发环境

1. Jupyter

• 可视化工具

1. Matplotlib、Seaborn、Plotly、Ggplot

• 计算引擎

1. NVIDIA GPUs、Google Cloud TPUs

• 代码笔记

1. GitHub、Kaggle、Colab

• 存储

1. S3、Google Cloud Storage (GCS)

• 可视化框架

1. TensorBoard、MLflow、Visdom、Neptune.ai

• 数据库

1. MySQL、MongoDB、PostgreSQL、SQLite

• 自动ML框架

1. Google Cloud AutoML

2. Azure Automated Machine Learning

3. Databricks AutoML、DataRobot AutoML

• 算法

1. Linear or Logistic Regression

2. Decision Trees or Random Forests

3. Gradient Boosting Machines (xgboost, lightgbm, etc)

4. Convolutional Neural Networks

5. Dense Neural Networks (MLPs, etc)

6. Bayesian Approaches

7. Recurrent Neural Networks

8. Generative Adversarial Networks

9. Evolutionary Approaches