This is a repo for ml workshop from college
Notes:
-----------------------Day 1-------------------
Teaching the computer without explicit code. make the machine understand.
Applications weather prediction stock market prediction You-tube video suggestion
Types supervised -provide past data -generate patterns -structure data(dog/cat analogy hair,size,etc) un-supervised -unstructured data
ONLY 5 different types of algorithms 1.classification 2.regression 3.anomaly detection 4.clustering 5.reinforcement
Classification(knn,decision tree classifier etc) knn - k nearest neighbor algorithm has discrete values
Regression has continuous values
Clustering group the data
Anomaly detection eg Spam changes in data
Reinforcement learning punishment/reward
Practical implementation step 1 -get data step2 -clean data step3 -make a model step4-train step5 -Predict
DEPENDENCIES PYTHON3 https://github.com/scriptonist/scripts
Apple/orange Gender classification packages scikitlearn -ml library, graph and stuff
================================== DAY 2===========================
references sirajology and udacity
knn algorithm implementation challenge: take 5 algorithm test with iris and find out the one with the most accuracy
70 accuracy is good
world challenge: create a face detection algorithm
Linear Regression
Random forest Give massive data, creates different decision tree and executes everything together
SDG classifier Schotiomatic gradient decent will optimise the the model to provide better accuracy and helps in generalisation Refer Jupyter Notebooks from the git repo of Udacity pytorch