Author Image

Hi, I am Tianxiao

Tianxiao He

CS PhD student at New York University

I am currently a PhD student in computer science at NYU in the field of computational neuroscience. I am a member of Neuroinformatics lab with Prof. Erdem Varol. My work is centered around developing machine learning tools to better analyze neural recordings. Specifically, my focus lies in processing electrophysiology data with downstream applications such as neural decoding.

Computational Neuroscience
Machine Learning
Neural Data Science

Education

New York University
2023-now
Ph.D in Computer Science
  • work in Prof. Erdem Varol's lab at Visualization and Data Analytics Research (VIDA) Center
  • received the School of Engineering (SoE) Fellowship from Department of Computer Science and Engineering
  • attend 11th Computational & Cognitive Neuroscience (CCN) Summer Program at Cold Spring Harbor Asia
Columbia University
2021-2023
B.S. in Computer Science
CGPA: 3.9 out of 4 (Dean's list)
  • worked as research assistant at Paninski lab at Columbia Zuckerman Institute from 2021-2023
  • presented a poster at Computational and Systems Neuroscience (COSYNE) conference 2023
  • led separate teams to work on projects on robotic learning, and hierarchical prediction with out of distribution problem
Taken Courses
Course NameTotal GPAObtained GPA
Artificial Intelligence44
Machine Learning43.75
Theoretical Neuroscience44
Computer Vision44
Bard College at Simon's Rock
2018-2021
B.A. in Computer Science
CGPA: 4 out of 4 (Dean's list)
  • worked as a teaching assistant for Python Programming, Algorithms & Data Structures
  • developed a carpooling web application for college students and implemented destination matching algorithms
  • participated in Hackathon and won the most technical award with a second-hand book exchange web application
  • developed an adventure game with Davinci Engine in Gamejam hosted by Lilith Studio (programming + plot design)

Research Experience

1
Density-based neural decoding for Neuropixel recordings
Paninski Lab @ Columbia University

Jun 2022 - Present, New York

Most decoding methods require spike sorting yet current spike sorting algorithms might be inaccurate and costly. So we proposed a spike sorting free method for decoding neural recordings.

Responsibilities:
  • Modeled the uncertainty of spike assignment using dynamical MoG for spike features distribution
  • Applied variational inference to fit the resulting model and to perform decoding
  • Benchmarked using recordings from various animals and probes shows better performance
  • Drafted and presented the poster at COSYNE 2023 conference
Analyze nonlinear embeddings of animal behavioral videos for decoding
Paninski Lab @ Columbia University

Jan 2022 - May 2022, New York

Animal behaviors are often studied with low dimensional task variables, but advances in computer vision enable us to study animal behavior with high-dimensional videos and apply it in decoding

Responsibilities:
  • Extracted nonlinear behavioral embeddings from animal videos using variational autoencoder
  • Decoded behavioral embeddings from electrophysiological recording with temporal convolution
  • Assessed decoder performance by comparing true behavior frames to reconstructed frames from predicted embeddings
2
3
Neural Network denoising for optogentically invoked signals
Paninski Lab @ Columbia University

Sep 2021 - Jan 2022, New York

Recent optogenetic advances allow us to stimulate multiple cells in quick succession. However, it is challenging to distinguish the postsynaptic current given overlapping trials from multiple cells.

Responsibilities:
  • Developed a NN temporal denoiser to demix current waveforms with spatio-temporal overlap
  • Ran a simulation study to test the model robustness to noise and sampling rate by synthesizing current traces under a variety of signal-to-noise regimes
  • Applied this denoising model in downstream tasks of mapping synaptic connections

Recent Projects

Card image cap
Grasping Robot Simulation

I implemented robot grasping tasks with Pytorch and Pybullet. I predicted grasp locations from image inputs with NN and developed a control policy for grasp execution and path planning.

April 7, 2022
Card image cap
Music Visualization

I designed music visualization algorithms, used the intermediate feature for genre classification and developed Android app for NN recognition of scanned images and playing its corresponding music.

April 28, 2021
Card image cap
Cultural Data Analysis

I analyzed World Value Survey data and reevaluated the idea of cultural group by clustering responses from different countries. I also examined traits of each cluster and their intercluster relationships.

January 3, 2021