Intersting links about deep learning
Dec 3, 2020·
·
1 min read
Dr. Georg Heiler
general
- http://jalammar.github.io/
- https://lilianweng.github.io/lil-log/
- http://colah.github.io/posts/2014-03-NN-Manifolds-Topology/
- fast.ai updated for 2020 an V2
- fastai book
- dive into deep learning
- Yes you should understand backprop
- CS 329S: Machine Learning Systems Design
critique
- Vincent Warmerdam: The profession of solving (the wrong problem) | PyData Amsterdam 2019
- A Recipe for Training Neural Networks
generic ML useful tools
constraints
graph neural networks
- https://medium.com/dair-ai/an-illustrated-guide-to-graph-neural-networks-d5564a551783
- [node classification with pytorch geometric] (https://colab.research.google.com/drive/14OvFnAXggxB8vM4e8vSURUp1TaKnovzX?usp=sharing#scrollTo=dszt2RUHE7lW)
frameworks
graphs
time-series
- sktime and sktime-dl
- pytorch-forecasting
- A Multivariate Deep Learning for Time Series Forecasting Library
- tsai state of the art sequence modeling
- GluonTS - Probabilistic Time Series Modeling in Python
- Awesome python time-series tooling not so much neural networks. But still very useful feature generators.
- ROCKET transforms time series using random convolutional kernels
- Orion - multivariate time-series anomaly detection
- How to prepare Time Series Data for LSTM Networks
- neural prophet
Varous time-series handling approaches:
- transformer https://www.youtube.com/watch?v=GiD87DeadYY
- 1D CNN, convert to images, convert to text https://www.youtube.com/watch?v=WoLlZLdoEQk
- Attention & time-series review https://towardsdatascience.com/attention-for-time-series-classification-and-forecasting-261723e0006d
- transformer:
- classificaton:
self supervised learing
anomaly detection
- Deep Few-shot Anomaly Detection https://towardsdatascience.com/deep-few-shot-anomaly-detection-b33f130d1f80
useful resources
tutorials
hardware

Authors
senior data expert
Georg is a Senior data expert at Magenta and a ML-ops engineer at ASCII.
He is solving challenges with data. His interests include geospatial graphs
and time series. Georg transitions the data platform of Magenta to the cloud
and is handling large scale multi-modal ML-ops challenges at ASCII.