Avatar

Amin Nejatbakhsh

Graduate Student of Theoretical Neuroscience

Columbia University

About Me

I am a graduate student of theoretical neuroscience at Columbia University. My research interests are quite broad, but I am mostly interested in modeling real-world physical data using tools and techniques from optimization, probabilistic modeling, machine learning, and dynamical systems. Apart from my academic life, I’m very much into sports and dancing. I play soccer and basketball very often and I sometimes swim, bike and run. I’m currently learning ballroom and latin dance and I’m part of a Persian dance group called GISU.

Interests

  • Neuroscience
  • Probability & Statistics
  • Dynamical Systems
  • Optimization
  • Machine Learning

Education

  • Ph.D. in Neuroscience, current

    Columbia University

  • M.A. and M.Ph. in Theoretical Neuroscience, 2019

    Columbia University

  • B.Sc. in Computer Engineering, Minor in Pure Mathematics, 2016

    Sharif University of Technology

Skills

Problem Solving

Statistics

Machine Learning

Dynamical Systems

Reinforcement Learning

MATLAB

Python

C and C++

LATEX

Experience

 
 
 
 
 

Graduate Student

Columbia University

Jan 2016 – Present New York City, US
What I’ve done so far includes:

  • Developing methods based on non-linear attractor reconstruciton for investigating causality in the networks of interacting nodes.
  • Developing methods for neuroscience data analysis using tools from optimal transport, statistical inference, matrix factorization, computer vision, machine learning, optimization, and dynamical systems.
 
 
 
 
 

Teaching Assistant

Columbia University

Jan 2016 – Present New York City, US
  • Fall 2020 - Foundations of Graphical Models
  • Spring 2020 - Theoretical Neuroscience
  • Spring 2020 - Math Tools for Neuroscientists (instructor & TA)
  • Fall 2019 - Probabilistic Programming Workshop
 
 
 
 
 

Software Developer

Hasin Co. & Torob

Jan 2012 – Jan 2014 Tehran, Iran
Responsibilities include:

  • Developing iOS Apps
  • Developing ML-Based Automated Clustering

Accomplish­ments

Student Travel Award, in MICCAI Conference, Peru

Gold Medal in International Mathematical Competition (IMC)

Recent Publications

Quickly discover relevant content by filtering publications.

Functional Causal Flow

Targeted manipulation of neural activity will be greatly facilitated by understanding causal interactions within neural ensembles. Here, we introduce a novel statistical method to infer a network’s “functional causal flow” (FCF) from ensemble neural recordings. Using ground truth data from models of cortical circuits, we show that FCF captures functional hierarchies in the ensemble and reliably predicts the effects of perturbing individual neurons or neural clusters. Critically, FCF is robust to noise and can be inferred from the activity of even a small fraction of neurons in the circuit. It thereby permits accurate prediction of circuit perturbation effects with existing recording technologies for the primate brain. We confirm this prediction by recording changes in the prefrontal ensemble spiking activity of alert monkeys in response to single-electrode microstimulation. Our results provide a foundation for using targeted circuit manipulations to develop new brain-machine interfaces or ameliorate cognitive dysfunctions in the human brain.

Probabilistic Joint Segmentation and Labeling of C. elegans Neurons

Automatic identification and segmentation of the neurons of C. elegans enables evaluating nervous system mutations, positional variability, and allows us to conduct high-throughput population studies employing many animals. A recently introduced transgene of C. elegans, named “NeuroPAL” has enabled the efficient annotation of neurons and the construction of a statistical atlas of their positions. Previous atlas-based segmentation approaches have modeled images of cells as a mixture model. The expectation-maximization (EM) algorithm and its variants are used to find the (local) maximum likelihood parameters for this class of models. We present a variation of the EM algorithm called Sinkhorn-EM (sEM) that uses regularized optimal transport Sinkhorn iterations to enforce constraints on the marginals of the joint distribution of observed variables and latent assignments in order to incorporate our prior information about cell sizes into the cluster-data assignment proportions. We apply our method to the problem of segmenting and labeling neurons in fluorescent microscopy images of C. elegans specimens. We show empirically that sEM outperforms vanilla EM and a recently proposed 3-step (filter, detect, identify) labeling approach. Open source code implementing this method is available at https://github.com/amin-nejat/SinkhornEM.

Deformable Non-negative Matrix Factorization

Extracting calcium traces from the neurons of mobile animals is a critical step in the study of the large-scale neuronal dynamics that govern behavior. Accurate activity extraction requires the correction of motion and movement-induced deformations as well as demixing of signals that may overlap spatially due to limitations in optical resolution. Traditionally, non-negative matrix factorization (NMF) methods have been successful in demixing and denoising cellular calcium activity in relatively motionless or pre-registered videos. However, standard NMF methods fail in mobile animals undergoing significant non-rigid motion; similarly, standard image registration methods based on template matching can fail when large changes in activity lead to mismatches with the image template. To address these issues simultaneously, we introduce a deformable non-negative matrix factorization (dNMF) framework that jointly optimizes registration with signal demixing. On simulated data and real C. elegans microscopy videos, dNMF outperforms traditional demixing methods that account for motion and demixing separately. Finally, following the extraction of neural traces from multiple imaging experiments, we develop a quantile regression time-series normalization technique to account for varying neural signal intensity baselines across different animals or different imaging setups.

Temporal Wasserstein Non-Negative Matrix Factorization

Motion segmentation for natural images commonly relies on dense optic flow to yield point trajectories which can be grouped into clusters through various means including spectral clustering or minimum cost multicuts. However, in biological imaging scenarios, such as fluorescence microscopy or calcium imaging, where the signal to noise ratio is compromised and intensity fluctuations occur, optical flow may be difficult to approximate. To this end, we propose an alternative paradigm for motion segmentation based on optimal transport which models the video frames as time-varying mass represented as histograms. Thus, we cast motion segmentation as a temporal non-linear matrix factorization problem with Wasserstein metric loss. The dictionary elements of this factorization yield segmentation of motion into coherent objects while the loading coefficients allow for time-varying intensity signal of the moving objects to be captured. We demonstrate the use of the proposed paradigm on a simulated multielectrode drift scenario, as well as calcium indicating fluorescence microscopy videos of the nematode Caenorhabditis elegans (C. elegans). The latter application has the added utility of extracting neural activity of the animal in freely conducted behavior.

NeuroPAL: a Neuronal Polychromatic Atlas of Landmarks for Whole Brain Imaging in C. elegans

Resolving whole-brain images of neuronal gene expression or neuronal activity patterns, to the level of single-neuron types with defined identities, represents a major challenge. We describe here the development and use of a multicolor Caenorhabditis elegans transgene, called “NeuroPAL” (a Neuronal Polychromatic Atlas of Landmarks), to resolve unique neural identities in whole-brain images. NeuroPAL worms share a stereotypical multicolor map, permitting complete, unambiguous and automated determination of individual neuron identities in conjunction with GCaMP-based neuronal activity reporters and GFP/YFP/CFP gene-expression reporters. To demonstrate the method and its potential, we use NeuroPAL and GFP-based reporters to map expression for the whole family of metabotropic acetylcholine, glutamate, and GABA neurotransmitter receptors encoded in the C. elegans genome, revealing a vast number of potential molecular connections that go far beyond the anatomically-defined connectome. We then expand the technique to whole-brain activity, employing NeuroPAL and a panneuronal neural-activity sensor (GCaMP6s) for functional analysis. Using this tool we delineate extensive nervous system activity patterns in response to several stimuli with single, identified neuron resolution. We find that attractive odors sensed by the same neuron class exhibit dissimilar activity patterns implying that, despite their shared valence and stimulus modality, these odors drive distinct neural circuitry. Our results also indicate that the connectome is a poor predictor of functional activity, emphasizing the need for comprehensive brain-activity recordings that delineate behavior-specific circuitry. Lastly, we illustrate the NeuroPAL as an unbiased analysis tool for investigating neuronal cell fate in specific mutant backgrounds. With these applications in mind, we establish a high-throughput software pipeline for automated and semi-automated cell identification using NeuroPAL. In conclusion, we demonstrate the power of the NeuroPAL as a tool for decoding whole-brain gene expression and maps of functional activity.

Robust Approximate Linear Regression Without Correspondence

Estimating regression coefficients using unordered multisets of covariates and responses has been introduced as the regression without correspondence problem. Previous theoretical analysis of the problem has been done in a setting where the responses are a permutation of the regressed covariates. This paper expands the setting by analyzing the problem where they may be missing correspondences and outliers in addition to a permutation action. We term this problem robust regression without correspondence and provide several algorithms for exact and approximate recovery in a noiseless and noisy one-dimensional setting as well as an approximation algorithm for multiple dimensions. The theoretical guarantees of the algorithms are verified in simulation data. We also demonstrate a neuroscience application by obtaining robust point set matchings of the neurons of the model organism Caenorhabditis elegans.

Projects

Gramophone

A native iOS app for downloading and listening to Iranian music.

3D Soccer Simulation

The RoboCup 3D Simulated Soccer League allows software agents to control humanoid robots to compete against one another in a realistic …

Advanced Numerical Calculator

A numerical calculator which allows users to perform advanced mathematical calculations and analysis such as integrating or …

Analyzing LFP Signals to Cluster V1 Neurons

Grating stimulus in multiple orientations were shown to the monkey while recording from its V1 layers. We implemented a system to …

Cinemas of India iOS Application

An iOS indian movie marketplace developed for both iPad and iPhone. Users can buy movies through this application and watch movies in …

Conversion Tool Between Regular Expressions and DFA/NFA Graphs

Both sides of conversion is available through the graphical user interface. Project for Theory of Languages and Automata course. …

ERP System

Object-oriented ERP system implemented based on UP methodology.

Hotel Reservation System

Web-based full front-end and back-end implementation of a hotel reservation system with smart recommender system.

Image Classifier

Feature extraction and classification system on corel image dataset

Implementation of a Voice Recognition System based on the Rat’s Auditory System

While the structure and dynamics of spiking neural networks (SNNs) may increase their computational power compared to traditional …

Intelligent 2048 Game Player

AI game player for 2048 game based on alpha/beta pruning of min-max state space graph.

Messaging System

Simple chat and messaging system.

Neuroscience Experimental Design and Data Collection System

In this project, I implemented an object oriented system with the aim of providing researchers the possibility to design visual or …

Population Dynamics Encoding Uncertainty and Reward in the Fronto Parietal Cortex

In this research, we are trying to find evidence that the brain encodes information about uncertainty, risk, and reward. We also want …

Search Engine

Crawler and search engine on Goodreads E-books.

Social Network

Full implementation of the front-end and back-end of a web-based social network (a simpler version of facebook).

Taaghche

Native iOS application for book reading.

Torob Automatic Feature Extraction and Clustering System

Torob is an intelligent shopping search engine which searches for a specific product through Iranian E-Stores. In order to find the …

Traffic Control and Vehicle Counting System

We used different image filters (gaussian filters, erosion and dilation), background subtraction methods (frame differences, gaussian …

Contact

  • 3227 Broadway, New York, NY, 10027, United States
  • Once entered the building take the elevators to the 5th floor
  • DM Me
  • Skype Me
  • Telegram Me