Last updated April 5, 2020

AI Physician, AI Music Composer & More Hot Releases This Week

Share

Published on April 4, 2020

by Ram Sagar

As the world came to a standstill in its fight against an invisible adversary, many researchers have almost ceased their operations. However, the ML community still has something to cheer for every week, with releases of new services that can be leveraged to build significant applications. From COVID-NET to an AWS based music composer, here are the top news for developers:

DeepMind RL Agent Outperforms Atari Benchmark

When it comes to creating new benchmarks for AGI, DeepMind keeps surprising the world with their innovations. After setting new benchmarks for GO by defeating human world champion, today DeepMind has announced that their reinforcement learning agent has scored above human baseline in Atari57.

The Arcade Learning Environment Atari57 was released in 2012 as an evaluation set of 57 classic Atari video games on which RL agents can be trained.

DeepMind’s Agent57 is the first deep reinforcement learning agent to master 57 Atari 2600 games. Agent57 combined an algorithm for efficient exploration with a meta-controller that adapts the exploration and long vs short-term behaviour of the agent. Adaptive meta-controller helps the agent to know when to explore and when to exploit, as well as what time-horizon it would be useful to learn with.

In 2012, DeepMind introduced its Deep Q-network (DQN) to tackle the Atari57 challenge, and since then, the AI research community has developed many extensions for and alternatives to the DQN.

Today, with Agent57, the researchers have succeeded in building a more generally intelligent agent that has above-human performance on all tasks in the Atari57 benchmark. The researchers are hopeful that this success can help a wide range of applications that can lead to AGI.

DarwinAI Open-Sources COVID-Net

In continued efforts to tackle COVID-19, DarwinAI, a Canadian startup has collaborated with researchers at the University of Waterloo and released COVID-Net— a convolutional neural network for COVID-19 detection via chest radiography.

The CEO, in his blog, wrote that they are open sourcing this model to the community with the hope of developing a robust tool to assist health care professionals in combating the pandemic. His team has also compiled together COVIDx, a dataset with 5941 chest radiography images across 2839 patient cases gathered from public sources.

Check the GitHub repo.

Buy AWS DeepComposer For $99

Amazon’s AWS has brought together their cloud advantage and the creative capabilities of generative adversarial networks(GANs) in the form of their service AWS DeepComposer, which can now be availed for just $99. With AWS DeepComposer, the users can train GAN models to create original music.

DeepComposercan gets hands-on, with a musical keyboard and the latest machine learning techniques, designed to expand your ML skills.

Developers, regardless of their background in ML or music, can also tweak the model hyperparameters and build custom GAN architecture with Amazon SageMaker.

Google’s AI Physician

Researchers at Google partnered with UCSF’s Bakar Computational Health Sciences Institute to develop a model, which evaluates the extent to which machine learning could anticipate standard prescribing patterns by doctors, based on electronic health records.

The dataset used for model training includes approximately three million medication orders from over 1,00,000 hospitalizations. The model uses electronic health record data that consists of names, addresses, contact details, record numbers, physician names, free-text notes, images, and more.

According to their experiments, the researchers have found the LSTM model to be the best. These models are capable of capturing the ordering and time recency of events in the data, making them a good choice for this problem.

Huawei Open-sources MindSpore

Huawei released their framework MindSpore to the machine learning community this week. MindSpore is designed to offer an all-scenario deep learning framework that best manifests the computing power of the Ascend AI processor while providing efficient execution for data scientists. Here are a few benefits: