Awesome

(Just one of the things I'm learning. https://github.com/hchiam/learning)

(even more ML stuff: https://github.com/hchiam/learning-ml)

machineLearning:

Simple test code for machine learning / neural networks / artificial intelligence (ML/NN/AI) in the Python programming language. And some live JavaScript examples too, like this one: https://codepen.io/hchiam/full/QGOyaE (for best results, open in Chrome).

No need to install a ton of things to import (more sophisticated code further down do need installations). Just read some commented code and get it running quickly to gather some intuitions.

You can have it even simpler and just run code in your browser without installing anything: here.

• neuralNet1.py: version 1 example of a very simplified neural network, using sensitivity parameter.
• neuralNet2.py: version 2 example of a very simplified neural network, with weighting based on "responsibilities" of different inputs. (This one seems really fast but may be unstable or naive because it's basically using learning sensitivity = 1.)
• neuralNet3.py: version 3 example of a very simplified neural network that combines version 1 and version 2, combining sensitivity parameter and "responsibilities" of different inputs.

• neuralNet4_Layered.py: version 4 example of a neural network that kinda combines version 2 and version 3, with learning error sensitivity, but also with a hidden layer. It also has a transformed version of the sigmoid function that goes from -1 to 1.

• predatorSim1D.py and predatorSim2D.py: animated simulations of a "predator" learning to move towards a target. Based on my turtle code and also a mix of my neural nets above: sensitivity parameter, weightings, inputs, but no hidden layer. See it run here or here.

machine learning web app:

You can try out the following web app live on CodePen: https://codepen.io/hchiam/full/rrwQRa.

Under "webApp_MachineLearning_Gesture" folder:

• gestures.html: the "structure" of the presentation of the web page.
• gestures.js: the "brains" of the web page. Tries to detect a mouse gesture when the mouse runs over the "pad".

Markov Word Generator - create words with the same "feel":

https://github.com/hchiam/word_gen

Notes and Code from Udacity course AI for Robotics:

https://github.com/hchiam/ai_for_robotics

Genetic Algorithm - applied to one of my linguistics projects:

https://github.com/hchiam/cogLang-geneticAlgo

Extra installation required but still pretty simple:

• neuralNet_iamtrask.py: "11 lines of code" for a 2-layer neural net code from http://iamtrask.github.io/2015/07/12/basic-python-network. Requires NumPy installed to run.

Google Developers:

The next few code samples are based on "Machine Learning Recipes with Josh Gordon" at: https://www.youtube.com/playlist?list=PLOU2XLYxmsIIuiBfYad6rFYQU_jL2ryal, which is also listed in the Google Developers YouTube channel.

• googleExample.py: machine learning in 6 lines of code, from https://youtu.be/cKxRvEZd3Mw, listed under the Google Developers channel on YouTube. Requires scikit-learn (sklearn) installed to run. Decision tree classifier. Supervised learning.

• googleExample2.py: decision tree classifier applied to Fisher's Iris flower data set, from https://www.youtube.com/watch?v=tNa99PG8hR8, listed under the Google Developers channel on YouTube. Requires scikit-learn and NumPy installed to run. You can also get a visualization (watch the video for how).

• googleExample3.py: a higher-level take on the decision tree classifier in googleExample2.py. Requires scikit-learn installed to run.

• googleExample4.py: building your our classifier. randomGuessClassifier() and KNearestNeighbours_barebones().

• googleExample5.py and googleExample6.md: image classification examples.

machinelearningmastery.com:

https://github.com/hchiam/machineLearning/blob/main/machinelearningmastery

Sirajology "Learn Python for Data Science" Challenges:

1. https://github.com/hchiam/gender_classification_challenge

Natural Evolution Strategies (NES) Example:

nes.py. See https://blog.openai.com/evolution-strategies/

Keras

https://github.com/hchiam/learning-keras

Synaptic.js

A JavaScript neural network library. My example codepen:

https://codepen.io/hchiam/pen/gWydZd?editors=1010

ml5.js web-friendly machine learning, built on TensorFlow.js

https://codepen.io/hchiam/pen/LrJVPQ

NLP with spaCy and textacy

https://github.com/hchiam/nlp_spacy_textacy

Learn more with freeCodeCamp:

For example, here's a video I found helpful for understanding RNNs and LSTM: https://www.freecodecamp.org/learn/machine-learning-with-python/how-neural-networks-work/recurrent-neural-networks-rnn-and-long-short-term-memory-lstm

LSTM: forget gate to forget irrelevant, input gate to remember relevant, and output gate to update new info

Then later reading up on attention and Transformers makes more sense.

Crash Course AI:

https://github.com/hchiam/crash-course-ai-labs

AutoML:

https://github.com/hchiam/learning-automl

Keep up to date:

https://www.youtube.com/@statquest - like this clear explanation of ROC and AUC or of transformers

https://www.youtube.com/@TwoMinutePapers

https://www.youtube.com/@twimlai - hear about things like AI-GAs, Quality-Diversity algorithms, jailbreaking, filters, adversarial training, pre-training, and more.

Links to more:

https://github.com/hchiam/learning-ml