Graph Deep Learning techniques on large graphs

Sometimes we encounter large graphs that force us beyond the available memory of our GPU or CPU. In these cases, we can utilize graph sampling techniques. PyTorch Geometric is a graph deep learning library that allows us to easily implement many graph neural network architectures with ease. The library contains many standard graph deep learning datasets like Cora, Citeseer, and Pubmed. But recently there’s been a push in graph open datasets to use large scale networks like the Open Graph Benchmark (OGB) [3]. In OGB, the various datasets range from ‘small’ networks like ogbn-arxiv (169,343 nodes) all the way up to ‘large’ datasets like ogbn-papers100M (111,059,956 nodes). …


The ISEE is the Independent School Entrance Exam and is required by many middle and high schools for the admissions process.

This guide goes through problems from the Middle Level Quantitative Reasoning section of the publicly available What to Expect on the ISEE practice test (scroll to the bottom of the pdf file to see the practice test), written by the ERB and found at isee.erblearn.org. …


The ISEE is the Independent School Entrance Exam and is required by many middle and high schools for the admissions process.

This guide goes through problems from the Lower Level Quantitative Reasoning section of the publicly available What to Expect on the ISEE practice test (scroll to the bottom of the pdf file to see the practice test), written by the ERB and found at isee.erblearn.org. …


The ISEE is the Independent School Entrance Exam and is required by many middle and high schools for the admissions process.

This guide goes through problems from the Upper Level Quantitative Reasoning section of the publicly available What to Expect on the ISEE practice test (scroll to the bottom of the pdf file to see the practice test), written by the ERB and found at isee.erblearn.org. …


My goal is to outline a lesson that any teacher can use in the classroom or any person interested in a very high level understanding of how AI works can walk through. This is not meant to be an exact representation of how AI truly works, but simply give intuition as to how it works. I have been a Math, SAT, ACT, ISEE tutor for close to a decade and work in machine learning research.

Pre-requisites: know what a probability is.

There are 2 sub-lessons, 1 smaller one and 1 larger one. …


My goal here is to create an easily adaptable framework to generate faces that look realistic, but also trick a facial recognition classifier. The example we will work through is the task of generating realistic faces that always classify as your face — despite them not being your face (or anyone’s face for that matter).

This is actually a tricky task because it involves updating the generator in two ways.

  1. Update the generator to make realistic images
  2. Update the generator adversarially to classify as your face

As you might expect, this will require two loss functions to update simultaneously. And if you thought updating a GAN was already a delicate procedure…you have not seen anything yet. …


Tricking facial recognition systems using adversarial attacks with GANs.

Image for post
Image for post

This is part of a series I am writing on tricking facial recognition systems using adversarial attacks with GANs.

However, before we trick a facial recognition classifier we need to build one to trick. I personally want to build one that can recognize my own face. Instead of training a neural network from scratch I can start with a pre-trained network and then finetune it to recognize my face. Finetuning is greatly beneficial as we can start with the model weights already trained on a large-scale face database and then update some of them to reflect the new tasks we want it to perform. These weights already understand how to recognize faces, but the only difference is it does not know my face. …


The goal of this post is to lay out a framework that could get you up and running with deep learning predictions on any dataframe using PyTorch and Pandas. By any dataframe I mean any combination of: categorical features, continuous features, datetime features, regression, binary classification, or multi-classification.

I may touch upon some of the technical aspects of what is going on behind the scenes, but mostly this is meant to be a framework discussion rather than a technical discussion. If you want to dig in further I suggest fast.ai courses in deep learning — and if you simply want predictions up and running without looking under the hood the fast.ai …


This is a continuation of a previous post where I do a full walkthrough of how to build an autonomous truck simulator using fast.ai, but ultimately these methods can work on any case where you need to finetune pretrained models or develop models that predict bounding boxes and classes together.

Now my goal is to walk through some of the more technical aspects of the training and inference processes and explain the details of how they are implemented in PyTorch. You can also reference the codebase in this Github repo.

Recall from the last post that there are two neural networks at work here. …


For some reason I had an inkling to go back to HMMs recently and here was the result. I started with this article from Nature by Sean R Eddy on biological sequence analysis and gene identification. Note: I am not a biologist and do not have any strong understanding of gene sequencing so it may (most certainly will) happen that I use the wrong terminology.

Here is an image from the article that lays out the general plan we will try to replicate.

Image for post
Image for post

The key here will be to try and identify when a 5' splice comes up as indicated in the State Path above. …

About

Mike Chaykowsky

Predictive Modeling. Based out of Los Angeles, CA LinkedIn: linkedin.com/in/chaykowsky

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store