Skip to main content

DEEP LEARNING




At this scale, a supercomputer would likely need terabytes of working memory just to store the model. The memory problem gets even worse when you bring GPUs into the picture. GPUs can process neural network workloads orders of magnitude faster than general purpose CPUs can, but each GPU has a relatively small amount of RAM—even the most expensive Nvidia Tesla GPUs only have 32GB of RAM. Medini says, "training such a model is prohibitive due to massive inter-GPU communication."

Instead of training on the entire 100 million outcomes—product purchases, in this example—Mach divides them into three "buckets," each containing 33.3 million randomly selected outcomes. Now, MACH creates another "world," and in that world, the 100 million outcomes are again randomly sorted into three buckets. Crucially, the random sorting is separate in World One and World Two—they each have the same 100 million outcomes, but their random distribution into buckets is different for each world.

With each world instantiated, a search is fed to both a "world one" classifier and a "world two" classifier, with only three possible outcomes apiece. "What is this person thinking about?" asks Shrivastava. "The most probable class is something that is common between these two buckets."

At this point, there are nine possible outcomes—three buckets in World One times three buckets in World Two. But MACH only needed to create six classes—World One's three buckets plus World Two's three buckets—to model that nine-outcome search space. This advantage improves as more "worlds" are created; a three-world approach produces 27 outcomes from only nine created classes, a four-world setup gives 81 outcomes from 12 classes, and so forth. "I am paying a cost linearly, and I am getting an exponential improvement," Shrivastava says.

Better yet, MACH lends itself better to distributed computing on smaller individual instances. The worlds "don't even have to talk to one another," Medini says. "In principle, you could train each [world] on a single GPU, which is something you could never do with a non-independent approach." In the real world, the researchers applied MACH to a 49 million product Amazon training database, randomly sorting it into 10,000 buckets in each of 32 separate worlds. That reduced the required parameters in the model more than an order of magnitude—and according to Medini, training the model required both less time and less memory than some of the best reported training times on models with comparable parameters.

Of course, this wouldn't be an Ars article on deep learning if we didn't close it out with a cynical reminder about unintended consequences. The unspoken reality is that the neural network isn't actually learning to show shoppers what they asked for. Instead, it's learning how to turn queries into purchases. The neural network doesn't know or care what the human was actually searching for; it just has an idea what that human is most likely to buy—and without sufficient oversight, systems trained to increase outcome probabilities this way can end up suggesting baby products to women who've suffered miscarriages, or worse.

Comments

Popular posts from this blog

Reliance launches e-commerce platform JioMart across several cities

           BY   DIVAKAR KUMAR PANDAY     Reliance has finally launched its e-commerce portal JioMart for several cities across India after running a pilot for limited pin codes in Mumbai, a move that is set to change the current positioning of the online grocery delivery service in India which is currently dominated by Bigbasket and Grofers. The service is live across Mumbai, Chennai, Kolkata, Hyderabad, Delhi, Bengaluru, Jaipur, Trivandrum and some tier II cities as well. During the pilot, JioMart was working with a WhatsApp business account. The JioMart website has listed essential grocery items as well as farm produce. The fruits and vegetables will be directly sourced from farmers that have collaborated with the brand. So technically, it would also compete with micro delivery platforms such as Swiggy’s Suprdaily, BBDaily and Milkbasket, and also farm to fork businesses including Ninjacart and Crofarm. The  platform  covers everything related to grocery, FMCG and consum

Nearly 2,000 e-contents uploaded to Magadh University website: VC

          BY  DIVAKAR KUMAR PANDAY. Ever since the introduction of online teaching in Magadh University in mid-April, nearly 2000 e-contents have been uploaded to the university website, says vice chancellor Prof Rajendra Prasad and fresh uploads are being made for the benefit of students. The good thing, according to the VC was that students of other universities too can benefit from the study material available on the MU website. The colleges too have been directed to provide links of their online teaching programmes on their respective websites. “Conceding that online teaching cannot be a substitute to classroom interactions,” the VC said that the university has to adapt itself to the post-covid realities. E-learning was here to stay post Covid as well in the form of supplement to class room teaching, said Prof Rajendra Prasad. Whereas the performance of PG departments and constituent colleges have been satisfactory in the field of online teaching, the same cannot be said about the

IIT Roorkee professor develops software to detect COVID19 in 5 second..

A professor at IIT-Roorkee has developed a software, which he claimed, can detect COVID-19 within five seconds. Kamal Jain IIT Roorkee’s Civil Engineering Department claimed that he took 40 days to develop the software, which will not only reduce time, but also testing costs drastically. At present, it takes about 24 to 48 hours to get COVID19 test results, and costs around Rs 4,500. The IIT professor also claimed that the new software will also reduce the risk of exposure to healthcare professionals. However, there has been no verification of his claim by any medical institution, including the Indian Council of Medical Research (ICMR). But, Kamal Jain is confident about his software, and has already filed a patent for the technique, and has also approached the ICMR for a endorsement. He claimed to have developed an AI-based database, which can analyze more than 60,000 X-ray scans, including those of COVID19, pneumonia and TB patients in about five seconds. In the pro