Most of the entities in the world are connected via any kind of relationship such as social networks, mobile networks, financial networks, user-product relationships. A graph (or network) is a natural way to represent such representations. Recently graph neural networks are getting paid a lot of attention from industries and academia to perform representation learning on graphs with neural networks so that we can use learned embedding for downstream tasks such as community detection, recommender systems, anomaly detection, and so forth. This talk covers an introduction to graph neural networks and their applications to various areas.
Advances in AI
Quantum computing has the potential to transform every domain, every business and every aspect of our lives, and the Quantum-AI combination is going to be a bonanza for business. Let’s understand how. Google’s quantum computer ‘Sycamore’ was been able to solve a complex mathematical calculation in just 3 minutes that a powerful supercomputer would have taken approximately 10,000 years. We call it ‘Quantum Supremacy’ in the world of ‘Qubits’ when a quantum computer outperforms a classic supercomputer. Now we can easily imagine, that with this enormous computational speed combined with the power of AI how it’s going to bring about a radical change in the way we perceive classic computing today.
A quantum computer is expected to be at least hundred million times faster than a classic computer. But it will be completely wrong if we think quantum computing will just bring extra speed because it will also bring some other significant dimensions specifically while combined with Machine Learning. Scientists and experts have already forecasted that Qubit in Quantum and Artificial Neurons in AI are going to essentially rule the scientific and technological arena of the future. We need to also understand that quantum computing is not just a boon but essentially the need of today and tomorrow because the silicon revolution is slowly collapsing as it has almost reached its limit. The rise of quantum along with AI is going to manifest the whole spectrum of new possibilities in the field of science and technology. Quantum computers can solve the wide range complex optimisation problems in each domain which conventional computing struggles to perform on time. These optimisation challenges are intrinsic in the field of finance, aviation, manufacturing, logistics, drug research and medicine, etc.
Light-weight Convolutional Neural Networks (CNNs) are mobile-friendly models that can provide inference without the need for any specialised hardware. These models can be very effective in point-of-care settings, where the detection of disease has to be performed in real-time. The talk will highlight few developments of the Medical Imaging Group (MIG) at the Indian Institute of Science, especially towards COVID19 management and diagnosis. Deployment of these lightweight networks on embedded platforms to show highly versatility as well as prove optimal performance in terms of being accurate will also be highlighted. The developed models having latency in the same order as other lightweight networks without compromising the accuracy will also be shown.
Science has fueled nearly every aspect of the world today, including healthcare, agriculture, transportation, entertainment, and the workplace. Judgement, as one of the core tenets of medicine, relies upon the integration of multi-layered data with nuanced decision making. I believe that we need to pay attention to it to make the best decisions at several crossroads of our lives. The thoughtful use of science can play a vital role in fixing the problems in society. At Rakuten, we attempt to treat cancer with light and develop AI algorithms to accelerate drug discovery, harness biomarkers to accurately match patients to clinical trials, and truly personalize cancer therapy using only patient’s own data. In this talk, we present our approaches towards conquering cancer using AI.
Visual metadata is a necessity to get videos discovered. The high cost of metadata makes it hard for the torso and tail to have good metadata – leading to a spiralling effect of the poor becoming poorer in the video domain. Our technology creates processes that make it extremely affordable to generate presentable metadata for an entire video library – almost automatically. A bottom-up video reasoning system automatically generates several structural artefacts, including poster art, trailers, cast information, promotions, brands, title credits, moods, and predict genre. All this is fed into a graph database with a global knowledge of TV, and from there on, curated by metadata professionals. We present an AI-assisted, metrics-driven process for creating and maintaining complete metadata and share experiences in taking this to industrial production.
Most modern AI systems are entirely neural. It is believed that for most tasks, end-to-end data-driven learning of neural architectures is sufficient, without any need for additional background knowledge about the domain. In this talk, I discuss the importance of symbolic AI in the modern neural world. With various applications ranging from natural language processing, planning, and computer vision, I will show the value of symbolic features, constraints, and representations inside deep learning systems.
Computer vision helps automating tasks at scale, by understanding relevant aspects of image or video content. In many cases scaling such tasks would be prohibitive, and they can now be carried at human-level accuracy. Applications include document processing for rapid user onboarding, product image classification and tagging, image quality control, and banner generation. This talk will provide an overview of vision research projects at the Rakuten Institute of Technology that help enabling frictionless customer experiences.
Natural Language Processing (NLP), as a field, has made dramatic progress over the last 10 years. In this talk, we will cover the underlying trends in data, compute and algorithms that are driving this progress. Diving into a bit more technical details, we will also look at some of the major deep-learning models that have resulted in fundamental advances in the field. Finally, we will wrap up with some applications of NLP to E-commerce at Rakuten.
