Machine Learning Weekly Review №7

mlreview.com — the source of latest credible papers, videos and projects on machine learning for scientists and engineers.


Recommended this week

Projects

  1. Mcmc-demo – interactive dashboard Deep Learning (DLSS) and Reinforcement Learning (RLSS) Summer School, Montreal 2017 various MCMC sampling algorithms.
  2. Kaldi — Speech Recognition Toolkit. Offers TensorFlow integration.
  3. Tensorflow and PyTorch implementations of 9 Generative Adversarial Networks (LSGAN, WGAN, DRAGAN, InfoGAN etc).
  4. AWD-LSTM — implementation of state-of-the-art Language Modeling in PyTorch. From Salesforce (Stephen Merity).
  5. Fashion-MNIST — A drop-in replacement of the original MNIST dataset.
  6. Nnabla — Neural Network Libraries by Sony. Python API, Dynamic computation graph, Multi-platform, Multi-GPU.
  7. Foolbox — Python toolbox to create adversarial examples that fool neural networks. Supports TF, PyTorch, MxNet etc.

Research Papers

  1. “Neural Machine Translation and Sequence-to-sequence Models: A Tutorial”. Does not assume any particular experience with neural networks or natural language processing from a reader. 65pp.
  2. Deep Learning for Video Game Playing” Survey by Niels Justesen. Reviews recent Deep Learning advances in a context of different types of games: shooter, strategy, arcade.
  3. Personalizing Session-based Recommendations with Hierarchical RNN” ACM RecSys 2017 paper by Massimo Quadrana with code.
  4. The prior can generally only be understood in the context of the likelihood”. Addresses paradox that although true prior should be independent of a model, in practice it’s usually motivated by likelihood.
  5. Reinforcement Learning with a Corrupted Reward Channel”. No real-world reward function is perfect and how to deal with it. By Tom Everitt, Victoria Krakovna, Laurent Orseau.
  6. TheoSea: Marching Theory to Light”. Can computer given VE figure out mathematically compact theories? TheoSea rediscovers Maxwell equations.
  7. UE4Sim: A Photo-Realistic Simulator for Computer Vision Applications”. Built on top of the Unreal Engine.
  8. Learning Transferable Architectures for Scalable Image Recognition”. ImageNet SOTA: 0.8% more and 9bln FLOPS less.
  9. “Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints”. EMNLP 2017 Award.

Posts, Articles, Tutorials

  1. Ensemble Learning to Improve Machine Learning Results”. How ensemble methods work: bagging, boosting and stacking. By Vadim Smolyakov.
  2. Background removal with deep learning” by Gidi Shperber.
  3. 10 Deep Learning projects based on Apache MXNet” by Julien Simon.
  4. Getting Inception Architectures to Work with Style Transfer” by Sahil Singla.
  5. An Alternative to Sgd: Stochastic Variance Reduction Methods”.
  6. How to Train a Simple Audio Recognition Network” in Tensorflow by Pete Warden.
  7. Tutorial — What is a variational autoencoder?”.
  8. Line-by-line Preprocessing Walkthrough for winning Data Science Bowl Submission by Brad Kenstler.
  9. Python For Finance: Algorithmic Trading” by Karlijn Willems.
  10. How does physics connect to machine learning?
  11. 5 Heroic Python NLP Libraries”: NLTK, TextBlob, Stanford CoreNLP, spaCy, gensim.
  12. StarCraft II RL Tutorial 1”. Deepmind’s StarCraft II RL Environment.

Free Books

  1. “Speech and Language Processing (3rd ed. draft)” — a leading textbook for natural language processing by Stanford Prof. Dan Jurafsky and Prof. James H. Martin.
  2. Computer Age Statistical Inference: Algorithms, Evidence and Data Science” by Stanford Prod. Bradley Efron.

Video Lectures and Talks

  1. Deep Learning for Self-Driving Cars” lectures and slides from MIT.
  2. “Pixel GAN autoencoder” talk by Alireza Makhzani. Slides.
  3. Deep Learning (DLSS) and Reinforcement Learning (RLSS) Summer School, Montreal 2017.
  4. JupyterCon Talks Recordings.
  5. Deep Learning for Natural Language Processing” Course by Oxford University. Features videos, slides and recommended reading.
  6. An Introduction to Deep Learning with TensorFlow” talk by Sebastian Raschka
  7. What is wrong with convolutional neural nets?” by Geoffrey Hinton at Fields Institute, 2017

You would also like

Facebook Group: Machine Learning Review