Series for Data science and Machine learning Noob to Pro

[rishikksh20]

It's always good to start from Scratch as discussed here https://github.com/pydatadelhi/talks/issues/43 :

1. Basic intuition behind machine learning. (Why it is necessary and basic algos and basic data pre-processing and analysis) I think this thing topic already cover but if necessary do it again.
2. Optimizing machine learning algos, Boosting, Bagging and Stacking ensemble learning (This topic is very necessary as it help you to win Kaggle competition as well as creating stable model) In real world scenarios all companies prefer ensemble model for stability reason.
3. Basic Learning Representation with Neural Network (Starting point for Deep learning) Beginning of new world for researchers.
4. Understanding of Neural Network (internal representation) and Optimisation Algos (How Neural network out perform traditional algos).
5. Convolution Neural network and image recognition problem
6. Sequence modelling and forecasting(Basic regression to simple markov models) Starting point for NLP and RNN.
7. Natural language Modelling and vector representation (Basic natural processing stuff)
8. Recurrent Neural network (LSTM,GRU) till this is enough for Machine learning.

   For data analysis and Data science concept we can arrange parallel tutorials for Pandas , Scikit-learn, Keras , Tensorflow etc and talk related to feature engineering, data cleaning, cross-validation etc. \ I am ready to take up any above's topic for talk, so I open for that.

Note: For everybody, feel free to jump in and make your suggestion in comments .

[manojpandey]

cc: @shagunsodhani @Dawny33

[manasRK]

Why forget the maths? I feel that basis Linear Algebra and Matrix Ops with Numpy/TF/PyTorch session would be useful as well. Would recommend first 5 chapters of Ian Goodfellow's book on DL.

Also, a session similar to @iamtrask's blog to deploy NNs from scratch is good as well. Plus some linguistics if you plan on doing NLP - would be a disservice otherwise. :)

[prabhant]

Hey Let's use this idea for giving PydataExpress tutorials in colleges(just like python express) to promote the upcoming Pydata Conference. So we can do few things Classify workshops in 3 categories: Beginner, Intermediate and Advanced. So that we can deliver the workshops according to the crowd. Though i agree that Maths is an important part of ML but I think we should stick to programming tutorials while delivering workshops. Though sticking to a single framework for these workshops will be ideal but that would limit the number of instructors, as some use keras others tensorflow and scikit learn. Some tutorials we can consider are Beginner:

• Building a classifier
• NLP/NLTK basics

Intermediate

• CNN, RNN
• Intro to Deep Learning

Advanced (I'm not quite sure what can be added in the advanced segment bit I think that workshops on implementation of research papers can be a part of it)

These workshops can be of minimum 1 hour to 2 hours Feel free to add suggest or add anything

[manojpandey]

This has a lot of good stuff https://github.com/hangtwenty/dive-into-machine-learning