AI In Graphics
Getting Intuition About Complex Math & More — AISeries — Episode #03
Hi, this post just presents a collection of images, Graph, and Gifs to Make You Grasp New Concepts Faster of Artificial Intelligence, ML, and DL ;)
Why visualizing? Visualizations allow us to access simultaneously a rich amount of information that can help us jump quickly to insights that may be hard to decipher from numerical calculations.
Einstein used visualization throughout his life. At age 16, Einstein used visualization when he discovered that the speed of light was always constant. Einstein believed that visual understanding was the most important form of education and more important than knowledge.
So Welcome! Enjoy!
01# First, What are LL EXPLORER LANDSCAPES?
This app allows you to interact with landscapes created with dimensionality reduction techniques and real data. The app has a variety of features including automatic gradient descent mode, live control of the descent rate (similar to the learning rate), the possibility of drawing gradient descent trajectories, studying gradient data, saving snapshots of any perspective in a variety of styles in png format, etc.
02# Second, What is Convolutional Neural Network?
Please, watch this awesome video:
03#Next, Convolutional Neural Networks for Visual Recognition.
Please visit this page from stanford.edu:
This network is running live in your browser, check it out!
04# how does the convolutional layer work?
There are the 3 key concepts of CNN (Convolutional neural networks):
2) Shared Weights and Biases
3) Activation & Pooling
Please refers to this link, look for Q&A#12:)
05#This demo trains a Convolutional Neural Network on the MNIST digits dataset in your browser right in front of your eyes:)
06#This is Andrej Karparthy toy 2d classification with a 2-layer neural network. Just awesome, Check it out!
07#what is the best algorithm for prediction in machine learning?
Machines are now able to learn from and train on their own by using previous computations and underlying algorithms to produce high-quality, easily reproducible decisions and results.
What follows are tailor-made machine learning algorithms that allow businesses to identify potential hazards and growth opportunities.
Here are some of the most commonly used algorithms for gradient-based optimization categorized on the basis of their type of machine learning:
08# what is the best framework for prediction in machine learning?
09#what is the best book to read for machine learning in python?
10#Is CPU or GPU more important for machine learning?
Getting Intuitive sense - CNN - Lectures by Brandon RohrerThe goal of this exploration was to get an intuitive sense for what types of functions and category boundaries neural networks can learn when used for regression or classification. We've seen both their power and their distinct preference for smoothness; We've only looked at two nonlinear activation functions, logistic and hyperbolic tangent, both of which are very closely related.There are lots of others and some of them do a bit better at capturing sharp nonlinearities.Rectified linear units or ReLUs, for instance, produce surfaces and boundaries that are quite a bit sharper. But my hope was to seed your intuition with some examples of what's actually going on under the hood when you train your neural network.Here are the most important things to walk away with:[First,] Neural networks learn functions and can be used for regression. Some activation functions limit the output range but as long as that matches the expected range of your outputs it's not a problem.Second, neural networks are most often used for classification; they've proven pretty good at it.Third, neural networks tend to create smooth functions when used for regression and smooth category boundaries when used for classification.Forth, for fully connected vanilla neural networks a two layer network can learn any function that a deep network can learn. However a deep network might be able to learn it with fewer nodes.Fifth, making sure that inputs are normalized that is they have a mean near zero and a standard deviation of less than one; this helps neural networks to be more sensitive to their relationships.I hope this helps you as you jump into your next project.Happy building!By Brandon Rohrer
@https://youtu.be/dPWYUELwIdM (49:40:00)
References & Credits
Andrej Karpathy blog
https://github.com/karpathy
https://twitter.com/karpathy
Deep Learning State of the Art (2020) https://youtu.be/0VH1Lim8gL8
How Deep Neural Networks Work by freeCodeCamp.org — Full Course for Beginners https://youtu.be/dPWYUELwIdM.
Deep Learning State of the Art (2020) by Lex Fridman —
Related Posts
00#Episode — AISeries — ML — Machine Learning Intro — What Is It and How It Evolves Over Time?
01#Episode — AISeries — Huawei ML FAQ — How do I get an HCIA certificate?
02#Episode — AISeries — Huawei ML FAQ Again — More annotation from Huawei Mock Exam
03#Episode — AISeries — AI In Graphics — Getting Intuition About Complex Math & More (this one:)
04#Episode — AISeries — Huawei ML FAQ — Advanced — Even More annotation from Huawei Mock Exam
05#Episode — AISeries — SVM — Credit Card — Start to Finished — A Complete Colab Notebook Using the Default of Credit Card Clients Data Set from UCI
06#Episode — AISeries — SVM — Breast Cancer — Start to Finished — A Complete Colab Notebook Using the Default of Credit Card Clients Data Set from UCI
07#Episode — AISeries — SVM — Cupcakes or Muffins? — Start To Finished — Based on Alice Zhao post
Online Demos
- AI Experiments With Goolge
- Convolutional Neural Network on MNIST digits
- Convolutional Neural Network on CIFAR-10
- Toy 2D data
- Toy 1D regression
- Training an Autoencoder on MNIST digits
- Deep Q Learning Reinforcement Learning demo
- Image Regression (“Painting”)
- Comparison of SGD/Adagrad/Adadelta on MNIST
The limits of deep learning are still on the process of being figured out!
By Lex Fridman