An all-too-common scenario: a seemingly impressive machine learning model is a complete failure when implemented in production. The fallout includes leaders who are now skeptical of machine learning and reluctant to try it again. One of the most likely culprits for this disconnect between results in development vs results in production is a poorly chosen validation set (or even worse, no validation set at all). Depending on the nature of your data, choosing a validation set can be the most important step. Although sklearn offers a train_test_split method, this method takes a random subset of the data, which is a poor choice for many real-world problems.
The shape-changing fins of fish are of great interest to engineers developing the locomotion and maneuvering capabilities of underwater and aerial systems. On page 310 of this issue, Pavlov et al. (1) find that the lymphatic circulatory system in tuna and other members of the Scombridae family of fish serves as a hydraulic system that actively changes the sweep angle of their second dorsal (back) and anal (underside) fins. This is an important finding because sweep-angle changes alter fin erectness and thereby affect the lift-force capacity of the fin.
The dye highlighted a large chamber at the base of the sickle-shaped median fins, and smaller channels nestled between the fish's back muscles and the fin also turned blue. It looked like the tuna could contract muscles at the base of the chamber to pressurize it, squeezing fluid out into the smaller channels and elevating its fin. So Block and her team analyzed high speed video of tuna schools in 20,000 gallon research tanks, looking to see if swimming tuna indeed raise and lower their fins at the angles that Pavlov experimentally observed. By looking at the lift and drag on a fish in 120 simulations of swimming, he found that raising those median fins stabilizes a tuna like a yacht keel--preventing roll-over.
What if a machine helped you make a dinner reservation? Quarles said AI helps optimize the search results when you want to make a reservation. For data compilation, OpenTable sees mobile devices as a big driver -- it's a rich data set to improve AI. "People don't want to spend more than two or three minutes [making a reservation], and they want to be confident they made the right decision," she said, suggesting that this large data set can make AI better and improve the results of search and the reservations.
A drone flying over a Florida beach Saturday caught on camera a struggle between a fisherman and a massive hammerhead shark. Curtis Williams, a drone operator, captured the shark fighting against the man. At certain points in the video, the man can be seen just inches away from the massive shark as a crowd gathers around. The Florida Fish and Wildlife Commission urges fishermen to "minimize fight time" with prohibited sharks by using a certain kind of tackle that helps to release the sharks.
A fisherman got one magnificent catch during the Fourth of July weekend when he reeled in a massive hammerhead shark off Florida's Panama City Beach -- and it was all captured in a drone video. Curtis Williams was flying his drone over Panama City Beach on Saturday when he saw the fisherman attempting to capture the shark. "Vacationing at beach and was going to video a nice pleasant sunset. The unidentified fisherman later released the shark back into the ocean.
Just as automation and machine learning have given internet companies detailed records and predictions of how users behave online, they can potentially enable scientists and government agencies to build similarly detailed models of the world's fisheries. "Today it's estimated that what's called illegal, unreported, and unregulated fishing costs the region between a half billion dollars and $1.5 billion a year," Zimring says. The Nature Conservancy is working with governments in the region, including in Palau, Micronesia, the Marshall Islands, and the Solomon Islands, to implement alternative monitoring programs, capturing video footage of fishing vessels instead of placing observers on each boat. To make analyzing the footage more feasible, The Nature Conservancy is investigating ways to use machine learning techniques to process that video material.
Deep learning models, trained by using a large set of labeled data and neural network architectures that contain many layers, routinely achieve impressive accuracy. The article explains how Neurala's technology will assist Air Shepherd's analysts in identifying animals, poachers, and vehicles from the terabytes of data created by the drones' video feeds. BeeScanning is a smartphone app that uses deep learning to analyze images of bee colonies to determine if they are infected by varroa mites. The Nature Conservancy's Indonesia Fisheries program is working with 2 technology companies that use machine learning to sort and recycle cell phones to develop a prototype, called Fishface, that applies this same technology to species identification for fish.
Google partnered with nonprofits to found Global Fishing Watch to detect illegal fishing activity using satellite data in near real time. The algorithm for detecting apparent fishing activity uses AIS data from roughly 35,000 fishing vessels worldwide (out of the roughly 200,000 vessels on the seas in the course of a year). The system captures the navigation pattern information of these 35,000 vessels with 22 million data points per day allowing deep learning algorithms to learn and then detect patterns that indicate fishing. Satellite intelligence startup Orbital Insight partnered with Global Forest Watch to detect illegal logging and other causes of global forest degradation.
Before you see data -- whether you are a baby learning a language or a scientist analyzing some data -- you start with a lot of uncertainty and then as you have more and more data you have more and more certainty," Ghahramani said. One of the center's area of study is trust and transparency around AI, while other areas of focus include policy, security and the impacts that AI could have on personhood. Adrian Weller, a senior researcher on Ghahramani's team and the trust and transparency research leader at the Leverhulme Centre for the Future of Intelligence, explained that AI systems based on machine learning use processes to arrive at decisions that do not mimic the "rational decision-making pathways" that humans comprehend. But by providing a means for making AI functions more transparent, commercial users of AI tools and their consumers could better understand how it works, determine its trustworthiness, and decide whether it is likely to meet the company's or its customer's needs.