|April 13 · Issue #36 · View online |
Hiya and welcome to an awesome week in deep learning!
| How Machine Learning May Help Tackle Depression | This study
published in Psychiatry Research shows that MRI scans can be analyzed by machine-learning algorithms (SVM) to establish the likelihood of someone suffering from the condition. As more labeled data becomes available training a ConvNet on this task would seem like a promising approach.
| OpenAI Just Beat Google DeepMind at Atari With an Algorithm From the 80s |
If you are looking for tendentious spin on the comparative performance of OpenAIs evolution strategies this article is for you.
| Exploring the mysteries of Go with AlphaGo and China's top players |
Mark you calendar, from May 23-27 the “Future of Go Summit” is taking place where AlphaGo will play against the world’s number one player, Ke Jie, as well as a team of China’s top pro players.
| Andrej Karpathy's Peek at Trends in Machine Learning |
Andrey Karpathy is using 5 years of data collected by his arxiv sanity
to explore interesting new trends in ML research.
| Unsupervised Sentiment Neuron |
A surprising result by the researchers at OpenAI who trained a large character predictor on Amazon review and found that simply predicting the next character in reviews resulted in discovering the concept of sentiment.
| Neural Network 3D Simulation |
A very cool 3D simulation of a neural network.
| Data Sciencing Motorcycles: Lean Assist |
A fun project using a convolutional neural network to classify images according to the current lean angle thereby, in theory, saving cyclist, motorized or not, from dragging their knees on the asphalt.
| Sonnet: TensorFlow-based Neural Network Library |
A new TensorFlow-based neural network library by DeepMind allowing developers to quickly build neural network modules with TF. The library uses an object-oriented approach, similar to Torch/NN, allowing modules to be created which define the forward pass of some computation.
| MNIST clock |
| NSynth: Neural Audio Synthesis |
Google’s Magenta team released a neural synthesizer to synthesize music and aid artist’s creative process. The release also includes a rich data set of musical notes that is an order of magnitude larger than other publicly available corpora.
| Google Research: An Open Source Sequence-to-Sequence Framework in TensorFlow |
Google releases a seq2seq framework in TensorFlow that Google Neural Machine Translation is based and which makes it easy to experiment with seq2seq models and achieve state-of-the-art results.
| One-Shot Video Object Segmentation |
Sequence of images demonstrating the object segmentation taken from the paper: https://arxiv.org/abs/1611.05198
| Neural Audio Synthesis of Musical Notes with WaveNet Autoencoders |
The paper accompanying Magenta’s NSynth release.
| Best Practices for Applying Deep Learning to Novel Applications |
A useful report containing actionable advice for domain experts wanting to apply deep learning.
We recently launched an events tab
on our website to compile all deep learning and machine learning events and give users an easy interface to search for those events by time and location. Upcoming events will also be featured in the newsletter. If we missed yours, just shoot us an email.
| 5th International Conference on Learning Representations |
ICLR takes a broad view of the field and includes topics such as deep learning and feature learning, metric learning, compositional modeling, reinforcement learning, and issues regarding large-scale learning and non-convex optimization.
| RE•WORK Deep Learning Summit Singapore (20% off for DL weekly subscribers) |
Join RE•WORK at the 11th Global Deep Learning Summit (http://bit.ly/2lniFrS) on 27-28 April and discover how Deep Learning, Machine Intelligence, and AI are positively impacting the world around us! Or attend the exclusive Finance track
focusing on Deep Learning applications in the financial sector. Use the discount code DLWEEKLY to get 20% off your passes!