What makes it the de-facto multi-class classifier loss function
There is much to learn and grasp to ace that important Data Science/Machine Learning interview. This article is part of a series of articles that try to make the preparation process easier and less daunting by introducing structure, use visual explanations, and keeping things relevant.
Motivation
The proper choice of loss functions is essential to train your machine learning models successfully. Thus questions about them are often brought up by the recruiters during the interview process. Questions may come from multiple fronts and angles, yet as long as you grasp the core concept, you will find that they won’t be much…
Why Deep Learning is so powerful yet so simple in its core
So you’ve heard about AI, heard about the amazing things a well-trained Machine Learning model, especially Deep Learning model can do. In some tasks, it even surpasses human performance. For example, a computer can now recognize different kinds of objects like cats, dogs, cars better, and an average human with faster speed, all thanks to the recent development of deep learning and neural networks. But what you may not hear about, is at its core, lies a simple theorem, a simple principle that makes all these possible. Enter Universal Approximation Theorem. Once you understand it, deep learning or multi-layer neural…
And How to Still Achieve Success
It All Started From My Personal Struggle
To tell you the truth, the fear of not being good enough at math has been hunting me from day one when I started my Machine Learning journey. It first struck me when I tried to apply the chain-rule while doing the back-propagation of neural networks. It took me a good several days before the assignment deadline until I finally figured out how to calculate the gradients. Man, that felt hard! I can almost still taste the bitterness and hardship in my mouth. Little did I know, this probably was the easy part. …
Came across this piece talking about the current #machinelearning status by Mark Saroufim. Some views are a bit controversial but good points nonetheless.
I’ll summarize the key points below: 🧵👇
- ML Researchers — Supposed to be risk-taking and less commercial oriented so ground-breaking progress can be made. Rather, ML Researchers found ways to not taking any risk but getting good pay through FANNG, media, YouTube, etc., and SOTA chasing.
- Math is overrated in Deep Learning. Matrix multiplication is mostly what you need and auto grad removes the real needs for manual gradient calculation. Be real now.
- The empiricism tendency of…
A Thought Experiment
What is UAT?
What exactly is Universal Approximation Theorem? Well, put in layman’s terms, UAT just means that giving a one hidden layer neural network with enough neurons. It can approximate(or simulate closely) any continuous function within the given input range. It means that a one hidden layer neural network is an ultimate flexible function approximator. Maybe a little too flexible.
A Lesson Learned
Because of the flexibility, Universal Approximation Theorem used to push AI researchers to focus mostly on shallow neural networks, thus in some way hinders the development progress of deep learning. This is interesting. Come to think of it, a ‘shallow and wide’…
Why All Data Scientists Not From Computer Science Background Should Take the Course
…my Journey into Data Science So Far
My Data Science self-education journey has always been a bit upside down. Instead of the usual Math > Computer Science > Language/Framework > Machine Learning > Deep Learning path. I took the other way around. I stumbled into Deep Learning from the fast.ai ‘Practical Deep Learning for Coders’ course first, which greatly boosted my interest and intention to learn more. Later I took the classic ‘Machine Learning’ course by Andrew Ng. (my review on these two courses here.) These are both great courses in their own merits and as I learn more and start doing my own projects, I found…
Data Scientist rejoice! Always have a hard time putting your fine LaTex equations into PowerPoint? Worry no more. Jeremy Howard from fast.ai to the rescue. Check out this great piece…
Good article from Kaspersky about Phishing Email detection using Machine Learning. The approach introduced are clearly laid out and easy to understand. Rarely found solid piece for ML/AI in cybersecurity…
Michael Li
Playing with numbers and shapes. Pushing features and pixels. Traversing multiple dimensions I sail on. Do I carry a towel you asked? The answer is 42 I smiled.
