Federated learning (FL) has emerged as a promising approach for training machine learning models across distributed data sources while preserving data privacy. However, FL faces significant challenges related to communication overhead and local resource constraints when balancing model requirements and communication capabilities. Particularly in the current era of large language models��
]]>
NVIDIA and the PyTorch team at Meta announced a groundbreaking collaboration that brings federated learning (FL) capabilities to mobile devices through the integration of NVIDIA FLARE and ExecuTorch. NVIDIA FLARE is a domain-agnostic, open-source, extensible SDK that enables researchers and data scientists to adapt existing machine learning or deep learning workflows to a federated paradigm.
]]>
For the past few months, the NVIDIA Collective Communications Library (NCCL) developers have been working hard on a set of new library features and bug fixes. In this post, we discuss the details of the NCCL 2.22 release and the pain points addressed. NVIDIA Magnum IO NCCL is a library designed to optimize inter-GPU and multi-node communication, crucial for efficient parallel computing��
]]>
The NVIDIA Mellanox Unified Fabric Manager (UFM) provides enhanced network management, monitoring, optimizations, security, and more. The newly announced UFM Cyber-AI applies AI to learning a data center��s operational cadence and network workload patterns, drawing on both real-time and historic telemetry and workload data. Against this baseline, it tracks the system��s health and network��
]]>
GTC Digital is all the great training, research, insights, and direct access to the brilliant minds of NVIDIA��s GPU Technology Conference, now online. Join live webinars, training, and Connect with the Experts sessions, or choose from a library of talks, panels, research posters, and demos that you can view on your own schedule, at your own pace. See all of the GTC Content here.
]]>
Australian-based TrademarkVision developed a deep learning-based reverse visual search platform that protects your brand by identifying similar trademarks from around the world. Simply upload your image to the platform, and their image recognition technology will compare it against other trademarked logos �C making it much easier to identify IP infringements than the previous time-consuming and��
]]>
Nuno Castro, the Director of Data Science at Expedia gave a talk at PyData London 2017 on how they��re using GPUs and deep learning to rank hotel images. Castro explains that humans first glance at the image within a hotel listing before considering the price or hotel name and this is why it��s crucial to rank images. Currently, the featured image of the listing on Expedia is manually selected and��
]]>
Researchers at Facebook Artificial Intelligence Research (FAIR) published a paper introducing AI-based dialog agents that can negotiate and compromise. In a new blog post, Facebook explains how existing chatbots can hold short conversations and perform simple tasks such as booking a restaurant �C but building machines that can hold meaningful conversations with people is challenging because it��
]]>
A Montreal-based startup developed a set of deep learning algorithms that can copy anyone��s voice with only 60 seconds of sample audio. Lyrebird, a startup spin-off from the MILA lab at University of Montr��al and advised by Aaron Courville and Yoshua Bengio claims to be the first of its kind to allow copying voices in a matter of minutes and control the emotion of the generation. Using CUDA��
]]>
Instacart, an Internet-based grocery delivery service, shares how they are using deep learning to help their tens of thousands personal shoppers be more efficient. ��By observing how our shoppers have picked millions of customer orders through our app, we have built models that predict the sequences our fastest shoppers will follow,�� mentions Instacart VP of Data Science Jeremy Stanly in a blog��
]]>
Artificial intelligence startup Sightengine developed image analysis technology to detect if images and videos contain offensive material, such as nudity, adult content or suggestive scenes. Using CUDA, Tesla K80s and cuDNN with the TensorFlow deep learning framework, the startup trained their deep learning models to determine whether or not an image is safe content and if it is appropriate to��
]]>
Engineers at Yelp, the popular crowd-sourced business review site, developed an algorithm that understands and determines which uploaded image should be a restaurant��s cover photo. In the past, cover photos were chosen by calculating a function based on votes, likes, upload date and image caption �C but the system was highly subject to selection bias. Cover photos are clicked much more than others��
]]>
California-based startup GoFindAI launched their new fashion app that makes shopping for clothes a breeze. Simply snap a photo of something you like, and the app��s AI-powered search engine scours over one million products from 1,000 online retailers looking for the same or similar items. Using CUDA and Tesla K80 GPUs on the Amazon and IBM SoftLayer clouds, the startup trained their deep learning��
]]>
Researchers from Google��s DeepMind and the University of Oxford developed a deep learning system that outperformed a professional lip reader. Using a TITAN X GPU, CUDA and the TensorFlow deep learning framework, the team trained their models on over 100,000 sentences from nearly 5,000 hours of BBC programs. By looking at each speaker��s lips, the system accurately deciphered entire phrases��
]]>
Citi Ventures made a strategic investment in Feedzai which uses deep learning to provide real-time fraud prevention in ecommerce and banking. Using CUDA, GTX 1080 GPUs and cuDNN with TensorFlow to train their deep learning models, Feedzai��s platform is able to scan large amounts of data to recognize evolving threats and then alerts customers in real-time to protect against fraud.
]]>
Home insurance quotes could become more accurate as a result of work by Palo Alto-based Cape Analytics, which raised $14 million to use computer vision and deep learning to improve automated property underwriting for insurance companies. The company brings geo-imagery from satellites into its cloud-based platform, and then runs the images through their deep learning algorithms that are able to��
]]>
The Barcelona-based startup developed a deep learning algorithm to determine in real-time what is in a particular image. RestB is commercializing their high-precision models by charging customers an initial training fee to build a custom model for their needs and charging per API call thereafter. ��If we take a picture of a city, (other company��s technology) would describe it by indicating that it��
]]>
The Russian machine learning firm Tselina Data Lab developed a deep learning-based camera algorithm called Fraudoscope that detects lies on facial emotions. Trained with CUDA and TITAN X GPUs, the lie-detecting app uses a high-definition camera to observe an interrogation and decode the results. The camera focuses on the interviewee �� the software maps changing pixels in the camera feed that��
]]>
A researcher from Carnegie Mellon University developed S.A.R.A. (Socially Aware Robot Assistant) that not only comprehends what you say, but also understands facial expressions and head movements. Using CUDA, GTX 1080 GPUs and cuDNN with TensorFlow to train the deep learning models, S.A.R.A. will reply differently if she detects a smile than someone frowning and offering a response that doesn��t��
]]>
BabbyCam is a new deep learning baby monitor that recognizes your baby, monitors their emotions and will alert you if their face is covered. As a new parent himself, the developer of the camera was in search for a solution with the ability to identify if the infant was on its stomach, one of the leading causes of Sudden Infant Death Syndrome (SIDS). With no monitors in existence, Benjamin Lui��
]]>
With the increasing volume of interactions with customers over social media channels, KLM Royal Dutch Airline is the first airline to test how artificial intelligence could assist customer service agents. ��We have 100,000 mentions a week on social media,�� says Tjalling Smit, senior vice president of Digital at KLM Royal Dutch Airlines. ��We handle around 15,000 customer service cases a week and we��
]]>
Announced at TwitchCon in San Diego, Logitech introduced their newest webcam designed to meet the needs for serious streamers. ��With the C922 Pro Stream Webcam, we are taking game casting to the next level so you can easily broadcast your gameplay like a pro, better engage with your community, and stand out from the crowd,�� said Guillaume Bourelly, portfolio and product manager of webcams at��
]]>
The popular digital radio platform iHeartRadio is using deep learning to help uncover what listeners find important when enjoying music. ��Humans perceive music through a wide variety of factors, including rhythm, instrumentation, tempo, and vocals,�� wrote Tim Schmeier, a data scientist at iHeartRadio in a recent blog. ��Some artists have a distinct ��sound�� while others are more varied.
]]>
Google released the latest version of their automatic image captioning model that is more accurate, and is much faster to train compared to the original system. ��The TensorFlow implementation released today achieves the same level of accuracy with significantly faster performance: time per training step is just 0.7 seconds in TensorFlow compared to 3 seconds in DistBelief (a system Google��
]]>
To make Siri great, Apple employed several artificial intelligence experts three years ago to apply deep learning to their intelligent mobile smart assistant. The team began training a neural net to replace the original Siri. ��We have the biggest and baddest GPU farm cranking all the time,�� says Alex Acero, who heads the speech team. ��The error rate has been cut by a factor of two in all the��
]]>
A new bot called Wonder will remember anything you want, and then return the information you need via a text message. Once you enter your phone number on the Wonder website, the bot will send you a text explaining how it works �C and then you just reply back with the information you want it to store. Using CUDA and GPUs in the Amazon cloud to train the deep learning models, the bot is able to��
]]>
With our planet getting warmer and warmer, and carbon dioxide levels steadily creeping up, companies are using deep learning to help cope with the effects that climate change is having on their crops. An article on MIT Technology Review highlights PEAT, a German company using CUDA, TITAN X GPUs and the cuDNN-accelerated Caffe deep learning framework to provide farmers with a plant disease and��
]]>
Ken Tran, Senior Research Engineer at Microsoft Research shares how they are using deep learning to create applications for people who are blind or visually impaired. Using Tesla M40 and TITAN X GPUs with the cuDNN-accelerated Caffe deep learning framework, they have trained their language model to describe images or scenes in natural language. The work is being used for Microsoft��s research��
]]>
Sometimes drawing a quick sketch is easier than using text to describe a particular product you have in mind. Researchers at Queen Mary University of London created a tool that recognizes hand-drawn sketches on a smartphone or tablet and uses them to search for products online. Using two Tesla K80 GPUs and the cuDNN-accelerated Caffe deep learning framework to train their convolutional neural��
]]>
Last Thursday at the International Conference on Machine Learning (ICML) in New York, Facebook announced a new piece of open source software aimed at streamlining and accelerating deep learning research. The software, named Torchnet, provides developers with a consistent set of widely used deep learning functions and utilities. Torchnet allows developers to write code in a consistent manner��
]]>
The popular social photo sharing site is adding the ability for people to take a photo of things in the real-world and get recommendations for similar products on Pinterest. For example, If you take a photo of a room with a purse and a lamp on a table that you like, Pinterest will show you results for all three items. Using TITAN X GPUs and the Caffe deep learning framework��
]]>
Dango is a new mobile that automatically suggests the best emoji, stickers and GIF images to include in your text conversations. The app relies on a neural network to understand the context of your conversation and predict the best visuals to help you communicate. To do this, the developers used the Theano deep learning framework and trained their models on hundreds of millions of real-world uses��
]]>
Maluuba, a Canadian artificial intelligence startup, created a system that can read, comprehend and reason, almost as well as humans. Their deep learning-based program called EpiReader is designed to solve a specific kind of comprehension task: a word is removed from a block of text and then the system is able to determine the missing word based on the context. Trained on TITAN X GPUs with the��
]]>
A new AI-powered device will be able to replace all of your various smart home control apps, as well as being able to recognize specific people and respond to a range of emotions and gestures. AI Build is a London-based startup focused on making your smart home more natural and intuitive. Powered by an NVIDIA Jetson TX1 and six cameras, the aiPort keeps track of your daily activities and uses��
]]>
Recommender systems are being deployed everywhere to deliver personalized experiences. Siraj Raval, a former software engineer at Meetup and CBS Interactive, recently launched an entertaining yet informative YouTube channel called Sirajology aimed to inspire and equip developers to build the future. His recent tutorial video explains how you can create a recommender system in just 10 lines of��
]]>
To meet the demand of the world��s growing population, farmers need to improve the productivity of their herds. Amsterdam-based Connecterra recently raised nearly $2 million to further develop their GPU-accelerated deep learning solution that consists of a wearable device that monitors each animal in the herd and transmits the data to a cloud platform for analysis and prediction of behavioral��
]]>
A new service called Bark will act as an online watchdog to protect your child from online predators and cyberbullying, as well as looking out for mental health concerns like depression or suicidal thoughts. Using deep learning and GPUs hosted in the Amazon Web Services cloud, Bark scans through social media messages and alerts parents of potential problems by flagging keywords and even analyzing��
]]>
With hundreds of millions of customers shopping daily on Amazon, it��s critical they help them discover the right product from their massive catalog of products. Amazon��s Deep Scalable Sparse Tensor Network Engine (DSSTNE or ��Destiny��) is a deep learning framework built from the ground up to help researchers develop search and recommendation systems. With multi-GPU support��
]]>
Imagine being able to virtually teleport from one space to another in real time. With 3D-capture cameras and a mixed reality display such as HoloLens, Microsoft Research��s new ��haloportation�� innovation allows users to see, hear, and interact with remote participants in 3D as if they are actually present in the same physical space. Custom software helps reconstruct every possibly viewpoint from��
]]>
Created by researchers at the Hong Kong University of Science and Technology, the MoodBox speaker is billed as the first ever high quality wireless speaker that senses human emotions. Using NVIDIA Tesla GPUs and deep learning, the speaker operates with cutting edge sensory recognition technology named ��Emi��. Emi collects and analyzes audio signals and music lyrics to provide efficient retrieval��
]]>
With over 150 million users and 10 billion monthly remote commands, Peel will begin using Baidu��s Deep Speech technology in their next-generation voice-enabled products. Deep Speech is a state-of-the-art GPU-trained speech recognition system developed using ��end-to-end deep learning�� by Baidu Research��s Silicon Valley AI Lab (SVAIL). Baidu��s Adam Coates, who leads the SVAIL team, said: ��Speech��
]]>
Yi-Jian Wu, Founder & CEO of Yuanqu Tech in China, talks about how NVIDIA Tesla GPUs are being used to train their interactive educational robot for children. Call the robot��s name and the speech-controlled robot is able to tell jokes, answer educational questions, teach English and act as a patient tutor for a child. For more information visit http://www.yuanqutech.com.
]]>
Today, Facebook introduced a new feature that automatically generates text descriptions of pictures using advanced object recognition technology. Until now, people using screen readers would only hear the name of the person who shared the photo, followed by the term ��photo�� when they came upon an image in News Feed. Now they will get a richer description of what��s in a photo. For instance��
]]>
United States call centers generate more than a billion hours of audio recordings every year and less than 25% of the audio is made searchable or analyzed. Launching out of Y Combinator��s Winter 2016 class, DeepGram uses deep learning and GPUs hosted in the Amazon Web Services cloud to quickly index audio and make it searchable. The free demo on the DeepGram site allows you to upload or provide��
]]>
After crossing 1.5M downloads, the free music-syncing app AmpMe now allows users to party without an internet connection using a new machine learning-powered predictive syncing technology. AmpMe streams music to multiple devices in perfect sync so friends can enjoy music together anywhere, without setting up a sound system. In the latest version, the new Predictive Sync�� feature allows AmpMe to��
]]>
Rob High, IBM Fellow, VP and chief technology officer for Watson, will deliver a keynote at the GPU Technology Conference in Silicon Valley on April 6. Watson was an overnight sensation when it competed on Jeopardy! against two former winners to win $1 million. Watson is among the first of a new generation of cognitive systems with far-reaching applications. It uses artificial intelligence��
]]>