Fei Xia

Mountain View, California, United States
8K followers 500+ connections

Join to view profile

About

Fei Xia is now a Senior Staff Software Engineer in Google, working on Generative AI to…

Activity

Join now to see all activity

Experience & Education

  • Google

View Fei’s full experience

See their title, tenure and more.

or

By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.

Licenses & Certifications

Publications

  • Block-coordinate Frank-Wolfe Optimization for Counting Objects in Images

    Neural Information Processing Systems (NeurIPS) Optimization Workshop

    We develop an optimization method to count objects in images. To avoid the detection of individual objects, which is computationally expensive and relies heavily on image quality, we model the object density as a linear transformation of each pixel feature, and obtain the object count by integrating density over the image. Learning such linear transformation is formulated as the minimization of a regularized quadratic function. Solving this optimization problem is highly nontrivial because it…

    We develop an optimization method to count objects in images. To avoid the detection of individual objects, which is computationally expensive and relies heavily on image quality, we model the object density as a linear transformation of each pixel feature, and obtain the object count by integrating density over the image. Learning such linear transformation is formulated as the minimization of a regularized quadratic function. Solving this optimization problem is highly nontrivial because it has exponentially large number of constraints. To cope with this challenge, inspired by the structural support vector machine (SVM), we explore the Block-Coordinate Frank-Wolfe (BCFW) algorithm, which is a state-of-the-art algorithm to solve structural SVM. However, BCFW cannot be directly applied to our problem. We derive the dual of our optimization problem and solve it by BCFW with modifications. Experiments show that BCFW solves our problem with lower iteration cost, faster convergence, and decent error rate.

    Other authors
    See publication

  • Max-margin Latent Feature Relational Models for Entity-Attribute Networks

    International Joint Conference on Neural Networks (IJCNN)

    Link prediction is a fundamental task in statistical analysis of network data. Though much research has concentrated on predicting entity-entity relationships in homogeneous networks, it has attracted increasing attentions to predict relationships in heterogeneous networks, which consist of multiple types of nodes and relational links. Existing work on heterogeneous network link prediction mainly focuses on using input features that are explicitly extracted by humans. This paper presents…

    Link prediction is a fundamental task in statistical analysis of network data. Though much research has concentrated on predicting entity-entity relationships in homogeneous networks, it has attracted increasing attentions to predict relationships in heterogeneous networks, which consist of multiple types of nodes and relational links. Existing work on heterogeneous network link prediction mainly focuses on using input features that are explicitly extracted by humans. This paper presents an
    approach to automatically learn latent features from partially observed heterogeneous networks, with a particular focus on entity-attribute networks (EANs), and making predictions for
    unseen pairs. To make the latent features discriminative, we adopt the max-margin idea under the framework of maximum entropy discrimination (MED). Our maximum entropy discrimination joint relational model (MED-JRM) can jointly predict entity-entity relationships as well as the missing attributes of entities in EANs. Experimental results on several real networks demonstrate
    that our model has improved performance over state-of-the-art homogeneous and heterogeneous network link prediction algorithms.

    Other authors
    See publication

  • Parameter Server for Distributed Machine Learning

    Neural Information Processing Systems (NeurIPS Workshop)

    We propose a parameter server framework to solve distributed machine learning problems. Both data and workload are distributed into client nodes, while server nodes maintain globally shared parameters, which are represented as sparse vectors and matrices. The framework manages asynchronous data communications between clients and servers. Flexible consistency models, elastic scalability and fault tolerance are supported by this framework. We present algorithms and theoretical analysis for…

    We propose a parameter server framework to solve distributed machine learning problems. Both data and workload are distributed into client nodes, while server nodes maintain globally shared parameters, which are represented as sparse vectors and matrices. The framework manages asynchronous data communications between clients and servers. Flexible consistency models, elastic scalability and fault tolerance are supported by this framework. We present algorithms and theoretical analysis for challenging nonconvex and nonsmooth problems. To demonstrate the scalability of the proposed framework, we show experimental results on real data with billions of parameters.

    Other authors
    See publication

  • Generalized Relational Topic Models with Data Augmentation

    International Joint Conferences on Artificial Intelligence (IJCAI)

    Relational topic models have shown promise on analyzing document network structures and discovering latent topic representations. This paper presents three extensions: 1) unlike the common link likelihood with a diagonal weight matrix that allows the-same-topic interactions only, we generalize it to use a full weight matrix that captures all pairwise topic interactions and is applicable to asymmetric networks; 2) instead of doing standard Bayesian inference, we perform regularized Bayesian…

    Relational topic models have shown promise on analyzing document network structures and discovering latent topic representations. This paper presents three extensions: 1) unlike the common link likelihood with a diagonal weight matrix that allows the-same-topic interactions only, we generalize it to use a full weight matrix that captures all pairwise topic interactions and is applicable to asymmetric networks; 2) instead of doing standard Bayesian inference, we perform regularized Bayesian inference with a regularization parameter to deal with the imbalanced link structure issue in common real networks; and 3) instead of doing variational approximation with strict mean-field assumptions, we present a collapsed Gibbs sampling algorithm for the generalized relational topic models without making restricting assumptions. Experimental results demonstrate the significance of these extensions on improving the prediction performance, and the time efficiency can be dramatically improved with a simple fast approximation method.

    Other authors
    See publication

  • Discriminative Relational Topic Models

    Submitted to Pattern Analysis and Machine Intelligence (PAMI)

    Many scientific and engineering fields involve analyzing network data. For document networks, relational topic models (RTMs) provide a probabilistic generative process to describe both the link structure and document contents, and they have shown promise on predicting network structures and discovering latent topic representations. However, existing RTMs have limitations in both the restricted model expressiveness and incapability of dealing with imbalanced network data. To expand the scope and…

    Many scientific and engineering fields involve analyzing network data. For document networks, relational topic models (RTMs) provide a probabilistic generative process to describe both the link structure and document contents, and they have shown promise on predicting network structures and discovering latent topic representations. However, existing RTMs have limitations in both the restricted model expressiveness and incapability of dealing with imbalanced network data. To expand the scope and improve the inference accuracy of RTMs, this paper presents three extensions: 1) unlike the common link likelihood with a diagonal weight matrix that allows the-same-topic interactions only, we generalize it to use a full weight matrix that captures all pairwise topic interactions and is applicable to asymmetric networks; 2) instead of doing standard Bayesian inference, we perform regularized Bayesian inference (RegBayes) with a regularization parameter to deal with the imbalanced link structure issue in common real networks and improve the discriminative ability of learned latent representations; and 3) instead of doing variational approximation with strict mean-field assumptions, we present collapsed Gibbs sampling algorithms for the generalized relational topic models by exploring data augmentation without making restricting assumptions. Under the generic RegBayes framework, we carefully investigate two popular discriminative loss functions, namely, the logistic log-loss and the max-margin hinge loss. Experimental results on several real network datasets demonstrate the significance of these extensions on improving the prediction performance, and the time efficiency can be dramatically improved with a simple fast approximation method.

    Other authors
    See publication

Projects

  • Location Recommendation Tool

    - Present

    1. Aim to deliver a geographic information analysis and recommendation tool for PNC Bank
    2. Mine information from Foursquare and Twitter data
    3. Develop a website under Django framework and with spatial database (PostgreSQL + PostGIS)

    Other creators
    • Rajarshi Das

  • Search Engine System

    -

    1. Implemented a search engine system with multiple retrieval algorithms, such as Boolean Retrieval model, BM25 and Indri, to retrieve Wikipedia documents indexed by LUCENE
    2. Optimized the system’s performance through using various techniques, such as multiple representation model, sequential dependency model and pseudo relevance feedback

Languages

  • Chinese

    Native or bilingual proficiency

  • English

    Full professional proficiency

More activity by Fei

View Fei’s full profile

  • See who you know in common
  • Get introduced
  • Contact Fei directly
Join to view full profile

Other similar profiles

Explore more posts

Explore top content on LinkedIn

Find curated posts and insights for relevant topics all in one place.

View top content

Add new skills with these courses

See all courses