A5069379894- Advanced Graph Neural Networks
- Complex Network Analysis Techniques
- Topic Modeling
- Domain Adaptation and Few-Shot Learning
- Graph Theory and Algorithms
- Generative Adversarial Networks and Image Synthesis
- Advanced Neural Network Applications
- Anomaly Detection Techniques and Applications
- Human Pose and Action Recognition
- Video Analysis and Summarization
- Functional Brain Connectivity Studies
- Multimodal Machine Learning Applications
- Spam and Phishing Detection
- Software Engineering Research
- Epigenetics and DNA Methylation
- Natural Language Processing Techniques
- Artificial Intelligence in Healthcare
- Advanced Image Processing Techniques
- Recommender Systems and Techniques
- Bioinformatics and Genomic Networks
- Data Mining Algorithms and Applications
- Advanced Image and Video Retrieval Techniques
- Misinformation and Its Impacts
- Machine Learning and Data Classification
- Brain Tumor Detection and Classification
Google (United States)
2016-2023
University of Southern California
2019-2021
Southern California University for Professional Studies
2020-2021
Integrated Systems Incorporation (United States)
2021
Stanford University
2013-2020
Many recent advancements in Computer Vision are attributed to large datasets. Open-source software packages for Machine Learning and inexpensive commodity hardware have reduced the barrier of entry exploring novel approaches at scale. It is possible train models over millions examples within a few days. Although large-scale datasets exist image understanding, such as ImageNet, there no comparable size video classification In this paper, we introduce YouTube-8M, largest multi-label dataset,...
Multi-person event recognition is a challenging task, often with many people active in the scene but only small subset contributing to an actual event. In this paper, we propose model which learns detect events such videos while automatically "attending" responsible for Our does not use explicit annotations regarding who or where those are during training and testing. particular, track recurrent neural network (RNN) represent features. We learn time-varying attention weights combine these...
Existing popular methods for semi-supervised learning with Graph Neural Networks (such as the Convolutional Network) provably cannot learn a general class of neighborhood mixing relationships. To address this weakness, we propose new model, MixHop, that can these relationships, including difference operators, by repeatedly feature representations neighbors at various distances. Mixhop requires no additional memory or computational complexity, and outperforms on challenging baselines. In...
There has been a surge of recent interest in learning representations for graph-structured data. Graph representation methods have generally fallen into three main categories, based on the availability labeled The first, network embedding (such as shallow graph or auto-encoders), focuses unsupervised relational structure. second, regularized neural networks, leverages graphs to augment losses with regularization objective semi-supervised learning. third, aims learn differentiable functions...
Graph embedding methods represent nodes in a continuous vector space, preserving information from the graph (e.g. by sampling random walks). There are many hyper-parameters to these (such as walk length) which have be manually tuned for every graph. In this paper, we replace with trainable parameters that automatically learn via backpropagation. particular, novel attention model on power series of transition matrix, guides optimize an upstream objective. Unlike previous approaches models,...
Interest surrounding cryptocurrencies, digital or virtual currencies that are used as a medium for financial transactions, has grown tremendously in the recent years. The anonymity these makes investors particularly susceptible to fraudity-such “pump and dump” scams-where goal is artificially inflate perceived worth of currency, luring victims into investing before fraudsters can sell their holdings. Because speed relative offered by social platforms such Twitter Telegram, media become...
We propose a new method for embedding graphs while preserving directed edge information. Learning such continuous-space vector representations (or embeddings) of nodes in graph is an important first step using network information (from social networks, user-item graphs, knowledge bases, etc.) many machine learning tasks. Unlike previous work, we (1) explicitly model as function node embeddings, and (2) novel objective, the likelihood, which contrasts from sampled random walks with...
The goal of video understanding is to develop algorithms that enable machines understand videos at the level human experts. Researchers have tackled various domains including classification, search, personalized recommendation, and more. However, there a research gap in combining these one unified learning framework. Towards that, we propose deep network embeds using their audio-visual content, onto metric space which preserves video-to-video relationships. Then, use trained embedding tackle...
Pre-trained convolutional neural networks (CNNs) are powerful off-the-shelf feature generators and have been shown to perform very well on a variety of tasks. Unfortunately, the generated features high dimensional expensive store: potentially hundreds thousands floats per example when processing videos. Traditional entropy based lossless compression methods little help as they do not yield desired level compression, while general purpose lossy energy compaction (e.g. PCA followed by...
Traditional recommendation systems using collaborative filtering (CF) approaches work relatively well when the candidate videos are sufficiently popular With increase of user-created videos, however, recommending fresh gets more and important, but pure CF-based may not perform in such cold-start situation. In this paper, we model as a video content-based similarity learning problem, learn deep embeddings trained to predict relationships identified by co-watch-based system only visual audial...
Graph Convolutional Networks (GCNs) have shown significant improvements in semi-supervised learning on graph-structured data. Concurrently, unsupervised of graph embeddings has benefited from the information contained random walks. In this paper, we propose a model: Network GCNs (N-GCN), which marries these two lines work. At its core, N-GCN trains multiple instances over node pairs discovered at different distances walks, and learns combination instance outputs optimizes classification...
Interest surrounding cryptocurrencies, digital or virtual currencies that are used as a medium for financial transactions, has grown tremendously in recent years. The anonymity these makes investors particularly susceptible to fraud---such ``pump and dump'' scams---where the goal is artificially inflate perceived worth of currency, luring victims into investing before fraudsters can sell their holdings. Because speed relative offered by social platforms such Twitter Telegram, media become...
TensorFlow-GNN (TF-GNN) is a scalable library for Graph Neural Networks in TensorFlow. It designed from the bottom up to support kinds of rich heterogeneous graph data that occurs today's information ecosystems. In addition enabling machine learning researchers and advanced developers, TF-GNN offers low-code solutions empower broader developer community learning. Many production models at Google use TF-GNN, it has been recently released as an open source project. this paper we describe...
Multi-person event recognition is a challenging task, often with many people active in the scene but only small subset contributing to an actual event. In this paper, we propose model which learns detect events such videos while automatically "attending" responsible for Our does not use explicit annotations regarding who or where those are during training and testing. particular, track recurrent neural network (RNN) represent features. We learn time-varying attention weights combine these...
In this paper we describe preliminary applications of network analysis techniques to eye-tracking data. a previous study, the first author conducted collaborative learning experiment in which subjects had access (or not) gaze-awareness tool: their task was learn from neuroscience diagrams remote collaboration. treatment group, they could see gaze partner displayed on screen real-time. control not. Dyads group achieved higher quality collaboration and gain. paper, how can further illuminate...
Precise hardware performance models play a crucial role in code optimizations. They can assist compilers making heuristic decisions or aid autotuners identifying the optimal configuration for given program. For example, autotuner XLA, machine learning compiler, discovered 10-20% speedup on state-of-the-art serving substantial production traffic at Google. Although there exist few datasets program prediction, they target small sub-programs such as basic blocks kernels. This paper introduces...