UpCode Academy Mobile Logo

Main Links

Signup Login

A Hacker's Guide to Practical Deep Learning

A Hacker's Guide to Practical Deep Learning featured image

About this Course

This course is targeted at individuals with some programming or data science background who wants to get started in the AI domain. We will would cover deep learning based AI techniques from first principles to help you get a grasp of how it works from the ground up. I will also cover various aspects of implementation, high level AI programming frameworks like Keras and Tensorflow as well as the basics how to tune an AI and improve the learning outcomes. I will also broadly cover the problems that deep learning is suited and also cover the topic of handwriting and specific image recognition using in-depth.

This course will be taught in Python 3.0 + other high level AI programming framework Attendees are expected to be able to program in Python and have a fair understanding of math and stats.


There is a lot of FOMO driving AI in every sector now, and a lot of hype surrounding AI. Many startups have rebranded themselves into one that is “AI-driven”, and Chinese and American investors are pouring in huge amounts of money into the space. On top of that, FANG and BAT are extremely avid in seeking acquisitions (e.g. DeepMind, DNNResearch), as well as going on technology demonstration sprees. Most notably, we have: Google’s AlphaGo beating Ke Jie, the world’s top Go player, and Baidu’s Xiaodu AI trumping humans in facial and voice recognition.

Everyone wants in on a piece of the action and a slice of the AI pie.

AI is definitely something you should at least have a sensing about because its impacts on each individual are definitely real. It will change our ways of life very drastically and reshape industries. It could be the best and worst of times. From self-driving cars to personal assistants, data analysis, and the creation of art, it will redefine the ways we live and learn, and it will bring with it many opportunities we

Start getting your hands dirty with code and data today!

Expectations & Goals

- Keras
- Tensorflow
- Forward propagation
- Backpropagation
- Pandas
- Matplotlib
- Num.py
- Activation functions: Sigmoid, Linear, ReLU
- Hyper parameters tuning
- Loss function
- Confusion matrix
- Stochastic gradient descent
- Data augmentation
- Overfitting / Underfitting
- Testing and Validation

What am I going to get from this course?

- Develop using iPython notebooks
- Understand AI use cases, development in the space & open questions in AI
- Implement a simple neural network from first principles
- Understand the problem domain of image recognition
- Work with networks likes CNN / ResNet
- Appreciate the intricacies and complexity that goes into deep learning algorithm
- Hands on in testing & validation
- Hands in tuning AI algorithms
- Manipulate data with num.py and pandas
- Visualize data with Matplotlib
- Set up a GPU environment

Required Materials

Laptop. The course will walk you through installing the necessary free software & there would be about an additional $250 spend on GPU compute environments.
Basic Programming Knowledge (Preferably Python 3.0).
At least high school level Math & Stats

Course Schedule

Lesson 1: Introductions to AI - April 14th, 2018 01:00PM - 03:30PM

- AI History + Use cases + Development of the space
- State of the Art Applications
- Open Questions in AI

Lesson 2: Deep Learning 101 - April 21st, 2018 01:00PM - 03:30PM

- Cats and dogs image recognition (code walkthrough)
- Simple 3x1 Neural in code network in code
- Exercise:
1. Implement simple neural network
2. Edit to make the network take 5 inputs and gives 2 outputs
3. Edit so that there are n number of middle layers

Lesson 3: Data exploration (Pandas + matplotlib PLT + numpy) - April 28th, 2018 01:00PM - 03:30PM

- Typical data set up & data type: CSV, logs, classes
- Why is data exploration and statistics important
- Hands on: Cats and dogs - histogram, scatter, image representation of RGB
- Setup Env for next lesson (Paperspace)

Lesson 4: Tensorflow (Code) - May 5th, 2018 01:00PM - 03:30PM

- Intro to tensorflow + CUDA
- Costliness in AI: Asking the right questions, approach, implementation, tuning
- Cats and dogs (base code) in tensorflow
- Exercise:
1. Change learning parameters
2. Segment data set
3. Implementation on Tensorflow 1

Lesson 5: Neural Network from First principles - May 12th, 2018 01:00PM - 03:30PM

- The loop of AI (Forward, backprop, epochs, etc)
- Activation functions - sigmoid, linear, ReLU
- Learning rates, Loss
- SGD as implementation of backpropagation

Lesson 6: Keras on top of Tensorflow - May 19th, 2018 01:00PM - 03:30PM

- Keras - high level abstraction | Why Keras or fast.ai
- Workflow
- Cats and dogs in Keras
- Exercise:
1. How to Improve learning
2. Read Keras docs and papers

Lesson 7: Validation & Testing (Code) - May 26th, 2018 01:00PM - 03:30PM

- Validation & Testing (Code)
- Data Augmentation
- Overfitting / Underfitting
- Classification by ANN is really an optimization problem
- Errors and confusion matrix
- Exercise:
1. Try to do OCR examples

Lesson 8: Computer vision examples & application - June 2nd, 2018 01:00PM - 03:30PM

- Computer vision examples & application
- ResNet on Cats & Dogs
- Keras code for MNIST Digits
- Where do you go from here?

Similar Courses