TensorFlow Weekly Issue 6

TensorFlow Weekly Issue 6 — August 28, 2018

Hi there, these are the top TensorFlow links from my weekly curation. If you're having trouble viewing this email, please read the online version.

TensorFlow Probability is a library for probabilistic reasoning and statistical analysis — TensorFlow Probability (TFP) is a Python library built on TensorFlow that makes it easy to combine probabilistic models and deep learning on modern hardware (TPU, GPU)...TFP includes: a) A wide selection of probability distributions and bijectors, b) Tools to build deep probabilistic models, including probabilistic layers and the Edward2 language, c) Variational inference and Markov chain Monte Carlo, d) Optimizers such as Nelder-Mead, BFGS, and SGLD, and more...Since TFP inherits the benefits of TensorFlow, you can build, fit, and deploy a model using a single language throughout the lifecycle of model exploration and production...
TensorFlow Team

Deep Learning OCR using TensorFlow and Python — In this post, deep learning neural networks are applied to the problem of optical character recognition (OCR) using Python and TensorFlow. This post makes use of TensorFlow and the convolutional neural network class available in the TFANN module.
Nicholas T Smith

Let’s build ‘Attention is all you need’ (Or why RNN s suck): part 1 of 2 — The most common neural networks used are called Long Short Term Memory (LSTM), but there are a variety based on the requirement like Gated Recurrent Units (GRU) which perform exceptionally well on sound models. RNNs have been very effective in solving the translation tasks, but there are numerous setbacks for using them...
Yash Bonde

Generating Music with Seq2Seq Models - Build a stacked LSTM encoder-decoder model with Keras — Sequence to sequence (Seq2Seq) or encoder-decoder models have shown to be extremely powerful for making translation engines and building chatbots. In this article, I will be building a encoder-decoder model that can learn to generate music from a bunch of midi files...I will be explaining the architecture of the model, the different components and steps involved and sharing my results and future work...I have structured the article to help someone new to deep learning to get an intuitive understanding on how to build a deep-learning model in Keras...
Noufal Samsudin

The future of model compression in 6 minutes: ‘DeepThin: A Self-Compressing Library for Deep Neural Networks' — Deep learning is expensive stuff...the industry is moving towards edge computing; predicting with models using a device’s local compute resources. Naturally, one wonders, “how will Apple put a PCIe slot in their iPhones?” but alas, I really mean local compute resources...To make this possible, model compression methods were devised — and today we’re going to look at the current SOA (state of the art) compression library: DeepThin...
Arya Vohra

How a badly configured Tensorflow in Docker can be 10x slower than expected — TensorFlow reads the number of logical CPU cores to configure itself, which can be all wrong when you have a container with CPU restriction...
Pierre Paci

Anaconda, Jupyter Notebook, TensorFlow and Keras for Deep Learning — So you want to get started to study deep learning? The first step is to set up the tools. In this post I will share with you how to set up Anaconda and Jupyter Notebook, and then install TensorFlow (including Keras)...
Margaret Maynard-Reid

Getting Data into TensorFlow Estimator Models — TensorFlow estimators work with input functions. The signature of an input function returns a tuple of features and labels. Features are a dictionary of feature names and numeric value arrays. Labels are an array of values. Some management needs to happen, such as shuffling the data, and returning it in batches. The approach you take determines how much effort you need to put in...
Robert John
 

Thanks for reading! If you have any feedback, just reply to this mail and I will get back to you.

Cheers,
Sebastian