Takuya Shimada

Takuya Shimada

Machine Learning Engineer

Preferred Networks, Inc., Japan

Biography

I am a machine learning engineer at Preferred Networks in Japan.

Interests

  • Weakly-supervised Learning
  • Representation Learning
  • Time Series Analysis

Education

  • MS in Computer Science, 2020

    The University of Tokyo, Japan

  • BS in Computer Science, 2018

    Waseda University, Japan

Experience

 
 
 
 
 

Machine Learning Engineer

Preferred Networks

Apr 2020 – Present Tokyo, Japan

Publications

Similarity-based Classification: Connecting Similarity Learning to Binary Classification

In real-world classification problems, pairwise supervision (i.e., a pair of patterns with a binary label indicating whether they …

Data interpolating prediction: Alternative interpretation of mixup

Data augmentation by mixing samples, such as Mixup, has widely been used typically for classification tasks. However, this strategy is …

Classification from pairwise similarities/dissimilarities and unlabeled data via empirical risk minimization

Pairwise similarities and dissimilarities between data points might be easier to obtain than fully labeled data in real-world …

Language-invariance in speaker adaption for zero resource scenario

Language-invariance of feature space maximum likelihood linear regression (fMLLR) is investigated in zero resource scenario. While …