The Ship List: MissingLink, May Edition
Summer is almost upon us, and we’re very excited to share with you what we’ve been working on this spring. We’ve developed a ton of new, powerful features to scale and accelerate your deep learning experiments, as well as new content and other updates. So, here we go…
New Features
Shared Data Storage
Train Models Faster and at Lower Cost
We’re excited to announce the release of the Shared Data Storage feature to streamline deep learning workflows for greater productivity. The shared data storage feature allows you to add a “parent” data volume (DV), an immutable data lake that manages your datasets and versions. This configuration enables multiple data volumes to reference a single copy of raw data, while each of them maintains its own index and metadata. This prevents the duplication of data, enhances productivity, and saves you significant storage costs.
Bi-Directional Links
Between Project Jobs and Experiments
We’ve made it easier and faster to track and navigate experiments using bi-directional links between jobs and experiments. We now have links from the experiments dashboard to jobs and vice versa in the experiment list, experiment info, jobs list, and job info pages.
Publications
MissingLink is powered by some of the most innovative minds in AI, who are constantly coming up with new ideas. Here’s some of the latest content from the MissingLink team and across the industry:
Most Common Neural Net PyTorch Mistakes
Last year, Andrej Karpathy, director of AI at Tesla, tweeted out PyTorch sage wisdom in 279 characters. This post covers each point Karpathy makes to see how these mistakes can manifest in a PyTorch code sample.
Train Models 6X Faster With Keras fit_generator Workers
Learn how Keras workers can make your experiments much faster in a snap. Auto-tracking experiments improves the recording of work which is valuable for an individual and an even bigger deal for teams. If you’re looking to try out workers or auto-tracking, make sure to:
Zero Shot Super Resolution Part 2: Implementing ZSSR with Keras and MissingLink.ai
In the second installment in our 4-part Zero Shot blog series, we review how to set up an environment for running the code we discussed in part 1: Single Image Super Resolution (SISR) methods and Zero-Shot Super Resolution.
Zero Shot Super Resolution Part 3: Training the Model
In the third installment in this series, we run through the code, discuss how it works, and highlight ways to modify it for future experiments.
Zencity + MissingLink: 20X More Experiments and 30% Lower Cloud Costs
We’ve produced a short video that shows how Zencity used MissingLink to go from managing deep learning via a laborious manual process to full automation of resources and experiments. In just a few short months, MissingLink helped Zencity:
- Accelerate training from 5 to 100 experiments per project
- Free up 30% of the data science team’s bandwidth spent on DevOps
- Save 30% of the cost of running experiments on Azure
Events
How AI is Making the World a Better Place
In an excellent write-up about the meetup we hosted at the MIT Media Lab in May, SamsungNEXT highlighted five of the innovative AI startups that showed up to demonstrate how they are making life healthier, safer — and cheaper!
- The team at Iterative Scopes presented how it’s improving the quality of colorectal cancer diagnosis, especially the ability to identify precancerous polyps.
- DeepCure showed how it’s reducing the cost of developing new drugs by using machine learning to figure out which molecules have the potential to treat various diseases.
- SmartVID.io talked about its machine learning-based video analysis software that analyzes construction site videos to identify unsafe practices and dangerous situations.
- Airworks discussed how it uses machine learning to increase the speed and drive down the cost and risk of construction jobs.
- Hopper showed how it is able to make predictions about flight and hotel prices — and save consumers a ton of money — using 15 trillion price quotes over a five year period reflecting the seasonality patterns within routes.,
Read the Post | Watch the full-length video
Next Up
Deep Learning in Healthcare
With successful experimental results and wide applications, deep learning has the potential to change the future of healthcare. The use of AI has become increasingly popular in the field, so it’s no surprise that a recent report noted that the AI healthcare market is expected to grow from $2.1 billion in 2018 to $36 billion by 2025. In this article, we review the many applications of deep learning in the healthcare industry, including electronic health records, cancer diagnosis, and disease prediction and treatment.
Coming Soon
We are busy working on more features and content to help you run faster, better deep learning experiments. So, stay tuned for updates here, on Twitter, Facebook, and LinkedIn. Or, let us show you what MissingLink can do for your team by requesting a demo.
MissingLink’s Boston Meetup was a great success, and we’re already busy planning the next one! Keep a lookout for new events centered around how AI is making the world a better place. Coming soon to venues in New York City and San Francisco.
MissingLink would be nowhere without the team and community that supports what we do. We want to thank these teams for providing their skills, knowledge, and resources: Hopper, DeepCure, AirWorks, SmartVid.io, CloudFactory, IterativeScopes and to MIT Media Lab for participating and hosting the event.
Originally published at https://missinglink.ai.
