Profile cover photo
Profile photo
Wayne Radinsky
18,602 followers -
Software Design Engineer
Software Design Engineer

18,602 followers
About
Posts

Post has attachment
"Artificial intelligence as a force for good." "With suicide prevention, every minute of response time matters. That's why the technology team at the well-known nonprofit Crisis Text Line in New York City analyzed some 65 million text messages to determine what words were most statistically associated with a high risk of suicide. This scale of analysis would clearly be infeasible without some form of automated analysis, and its results surprised the team. Use of the term 'EMS' in a text, for example, is five times more predictive of a high risk of suicide than the actual word 'suicide.' By using this analysis, the team can now better prioritize incoming messages, much like the triage system in a hospital emergency department. As a result, the organization is now able to respond to 94 percent of high-risk texters in fewer than five minutes. This is just one example of 'mission-driven artificial intelligence'."
Add a comment...

Post has attachment
"If you are a gamer, you must have heard of the two insanely popular Battle Royale games out right now, Fortnite and PUBG. They are two very similar games in which 100 players duke it out on a small island until there is just one survivor remaining. I like the gameplay of Fortnite but tend to prefer the more realistic visuals of PUBG. This got me thinking, can we have graphics mods for games that can allow us to choose the visual effects of our liking without having to rely on the game developers providing us that option? What if a mod was available that could render the frames of Fortnite in the visuals of PUBG? That's where I decided to explore if Deep Learning could help and I came across neural networks called CycleGANs that happen to be very good at image style transfer. In this article, I'll go over how CycleGANs work and then train them to visually convert Fortnite into PUBG."
Add a comment...

Post has attachment
"When used to test three of top performing DNNs from the Udacity self-driving car challenge, it unearthed thousands of erroneous behaviours, many of which could lead to potentially fatal crashes."

"We can't apply traditional measures such as statement coverage to understand how well tested a DNN is, so we need to find an alternate metric. Borrowing from DeepXplore, DeepTest uses a notion of neuron coverage. Another interesting challenge is how you know whether the output of the model is correct in any given scenario. DeepXplore introduced the notion of using an ensemble of models to detect models making unusual predictions for a given input. DeepTest has a neat twist on this, using an ensemble of inputs which should all lead to the same output (e.g. the same road in different weather and visibility conditions) to detect erroneous outputs."
Add a comment...

Post has attachment
"Naftali Tishby of the Hebrew University of Jerusalem believes the training processes in neural networks illustrate a branch of information theory that he helped develop two decades ago. He coined the term 'information bottleneck' to describe the most efficient way that a system can find relationships between only the pieces of data that matter for a particular task and treat everything else within the sample as irrelevant noise."

"We believe [Tishby's] ideas are substantially correct, but there are a few technical details that have to be worked out. The fact that we converged to similar ideas is remarkable because we started from completely independent premises."

"Soatto and his UCLA colleague Alessandro Achille used ideas from the information bottleneck to develop training optimizations that help smaller networks tune out noise."
Add a comment...

Post has attachment
"China is winning the global tech race." Of the top 50 entries of the Wikipedia ranking of unicorn start-ups by value, 26 are Chinese, 16 are from the US, and none are from Europe.
Add a comment...

Post has attachment
"Various deep learning based algorithms have been proposed to aid pathologists in effectively reviewing pathology slides and detecting cancer metastasis. Because of the outrageously large size of the original digital slides, most of the algorithms currently being used split the slide into lots of smaller individual image patches, e.g. 256x256 pixels. A deep convolutional neural network is then trained to classify whether each small patch contains tumor cells or normal cells separately. However, sometimes it is difficult to predict whether a patch contains tumor cells without knowing its surroundings, especially around the tumor/normal boundary regions, and false positive predictions are often introduced. Figure 2 shows one example of how difficult this can be."

"We have proposed a new deep learning algorithm that takes not just one individual patch but a grid of neighboring patches as input to jointly predict whether they are tumor cells or normal cells."
Baidu Research
Baidu Research
research.baidu.com
Add a comment...

Post has attachment
"How Hulu builds its industry leading recommendation engine." Their system is called CF-NADE, which means "neural autoregressive distribution estimator for collaborative filtering."

"CF-NADE models the distribution of the vector of user ratings, optimized by maximizing the joint probability of all the vectors. Let's take an example. Suppose a user rated 4 movies, 'Transformers', 'SpongeBob', 'Teenage Mutant Ninja Turtles' and 'Interstellar', with scores 4,2,3 and 5, respectively, on a 5-star scale. ..."

"In practice, explicit feedback is rare, but implicit feedback like watch/browse/search/purchase behaviors are abundant. Adapting CF-NADE to implicit feedback will be our focus in near future."
Add a comment...

Post has attachment
Model Zoo: Discover open source deep learning code and pretrained models. Has models for TensorFlow, Keras, PyTorth, Caffe, Caffe2, and MXNet, categorized into computer vision, natural language processing, audio and speech, generative models, and reinforcement learning.
Add a comment...

Post has attachment
SLM Lab is a modular deep reinforcement learning framework in PyTorch.
kengz/SLM-Lab
kengz/SLM-Lab
github.com
Add a comment...

Post has attachment
"PyTorch Geometric is a geometric deep learning extension library for PyTorch. It consists of various methods for deep learning on graphs and other irregular structures, also known as geometric deep learning, from a variety of published papers."

"The following methods are currently implemented: SplineConv, GCNConv, ChebConv, NNConv, GATConv, AGNNProp, SAGEConv, Graclus Pooling, and Voxel Grid Pooling."
Add a comment...
Wait while more posts are being loaded