Scribd Logo
Upload

Welcome to Scribd!

  • 42 views

    Tell Me About Your Yourself

    Uploaded by

    Madi Shimizu
    The document discusses the applicant's background and interests. As a data scientist, they analyze EHR data using prediction and extraction models to solve medical problems. As a machine learning researcher, their work includes adverse drug event extraction, medical NLI, computer vision, and physics-informed machine learning. They have a bachelor's in electronics engineering and a master's in software systems. They are interested in pursuing a PhD to systematically learn Bayesian modeling and have industry experience with medical chatbots and question answering startups. They are interested in the program due to its differential privacy and DNA methylation research.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Tell Me About Your Yourself

    Uploaded by

    Madi Shimizu
    42 views3 pages
    The document discusses the applicant's background and interests. As a data scientist, they analyze EHR data using prediction and extraction models to solve medical problems. As a machine learning researcher, their work includes adverse drug event extraction, medical NLI, computer vision, and physics-informed machine learning. They have a bachelor's in electronics engineering and a master's in software systems. They are interested in pursuing a PhD to systematically learn Bayesian modeling and have industry experience with medical chatbots and question answering startups. They are interested in the program due to its differential privacy and DNA methylation research.

    Original Title

    interview-computational-biology

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses the applicant's background and interests. As a data scientist, they analyze EHR data using prediction and extraction models to solve medical problems. As a machine learning researcher, their work includes adverse drug event extraction, medical NLI, computer vision, and physics-informed machine learning. They have a bachelor's in electronics engineering and a master's in software systems. They are interested in pursuing a PhD to systematically learn Bayesian modeling and have industry experience with medical chatbots and question answering startups. They are interested in the program due to its differential privacy and DNA methylation research.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    42 views3 pages

    Tell Me About Your Yourself

    Uploaded by

    Madi Shimizu
    The document discusses the applicant's background and interests. As a data scientist, they analyze EHR data using prediction and extraction models to solve medical problems. As a machine learning researcher, their work includes adverse drug event extraction, medical NLI, computer vision, and physics-informed machine learning. They have a bachelor's in electronics engineering and a master's in software systems. They are interested in pursuing a PhD to systematically learn Bayesian modeling and have industry experience with medical chatbots and question answering startups. They are interested in the program due to its differential privacy and DNA methylation research.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    Tell me about your yourself

    I'm a data scientist.


    What does a data scientist do?
    We analyze data and build models to solve problems
    What kind of data?
    EHR data

    Visits
    Events
    Treatment
    Medical History
    Structured Data
    What kind of models?
    Prediction of an event (cardiac failure)
    Adverse Drug event extraction
    I'm a Machine Learning Researcher

    Adverse Drug Event Extraction paper


    Med NLI : Hypothesis is true or false or undetermined given a premise
    Computer Vision (Masked RCNN) : Parking lot detected from Satellite images
    Bayesian Modeling of Bose-Einstein Condensate

    5th state of matter (gas of bosons cooled to absolute zero -273.15 C)


    Wave-like behaviour
    Modeling Wave function using Gaussian Processes from noisy observations generated by a simulator
    Physics-informed Machine Learning
    Data-driven solutions to PDE

    boundary conditions
    noisy observations
    Data-driven discovery
    Automatically build mathematical models from data
    Bachelor's in Electronics and Instrumentation Engineering
    Robotics club
    Raspberry Pi, Arduino, Micro-controllers
    Programming a MC and building circuitry
    Linux kernel
    Master's in Software Systems
    Why? Linux kernel
    Wireless Sensor Networks
    Routing protocols
    Time synchronization
    Energy Conservation
    Exploration algorithm using Linear Programming
    Break
    Talk by Prof. C.V. Jawahar
    Convolutional Neural Network for Computer Vision
    Automatic Representation Learning
    I saw the potential in it
    Wrote blogs on Neural Network based Chatbots (seq2seq)
    Probabilistic Perspective
    Joined FSF & FSHM
    Traveled with them
    Take classes on Python and Machine Learning at Summer Camp and Winter Camps
    Talks and Workshops at conferences (4CCON & Startups)
    Startups
    voicy.ai
    Chatbots
    Java backend
    Reinforcement learning for Hybrid Code Networks
    datalog.ai
    Question Answering (Reading Comprehension)
    High-capacity models Tensorflow
    Model-parallelism, Data-parallelism

    Why do you want to do a PhD?


    Building a bayesian model is an art (so far). Can we transform it into science?
    I see the same potential that I saw in CNNs back in 2015.
    That changed my life in 2015
    Something that is worthy of my time and effort
    What I have learned seems superficial.
    Best way to learn the art of Bayesian Modeling is by doing it practically
    I need a platform where I can apply it and learn from the experience
    I need a support system
    I need to document my experience

    Why this program?


    What

    What do you know about our research?


    Differential Privacy
    Clipping and perturbing gradients
    DNA Methylation
    Addition of Methyl CH3- to cytosine
    CpG dinucleotide (phosphate)
    Most somatic cells, CpG are methylated
    Unmethylated CpG : CpG islands (promoters in genes)
    Differentially Methylated Regions (DMR)

    DNA sequences with varying methylation profiles


    Aberrant DMR patterns : auto-immune, metabolic, psychological disorders
    Cancer
    Genomic Analysis : Quantifying differences in DNA methylation across large number of samples
    Effects on regulating Gene Expression
    Distinguishing DMR patterns between different cancer types on the whole genomic level
    Data from TCGA (The Cancer Genomic Atlas)
    10000 profiles, 33 kinds of cancer
    220 x 1663 matrix
    220 CpG sites, 1663 beta values
    93% accuracy
    Feature Visualization : which CpG site led to the cancer

    You might also like