Deep Learning
This is a course for data scientists, data analysts, engineers, researchers, software developers, and quants.
Prerequisites
Prior experience with Python. High-school math knowledge is recommended. A quantitative background and familiarity with basic probability and linear algebra would also be beneficial but are not required.
You may skip day 1 if you have recently completed either of Python Charmers' courses Python for Predictive Data Analytics or Python for Scientists and Engineers.
Expected Outcomes
This course introduces machine learning and deep learning using the Python packages scikit-learn and PyTorch
By the end of the course, you will understand the concepts of classical ML algorithms as well as neural networks, convolutional neural networks, and transformers, and you will have experience applying these in practice to develop and refine models for classification and regression across various domains.
Course Syllabus
Session 1: Machine Learning
Day 1 gives you a practical and comprehensive introduction to machine learning for powerfully inferring complex models from data, with examples from a range of industries, including time-series and spatial datasets:
- Crash course in NumPy
- Concepts in machine learning (ML)
- Overview of the ML package ecosystem in Python
- Regression and classification with scikit-learn, with applications: time-series, diagnosis, image recognition, satellite, imagery, ...
- Validation and model selection; diagnostic tools; yellowbrick
- Feature engineering and selection; eli5, text analysis
- Outlier and anomaly detection with pyOD
Session 2: Deep Learning
Day 2 introduces the approach to machine learning known as “deep learning”, using neural networks trained with the PyTorch library on GPUs:
- PyTorch nuts and bolts; using GPUs
- Concepts in deep learning (DL): network architectures, activation functions, loss functions, SGD
- Similarities and differences of PyTorch vs scikit-learn; the skorch API
- Residual and convolutional networks for time-series and image data
- Regularization: weight decay, dropout, pooling, batch normalization
Session 3: Current and Future Directions; Best Practices
Day 3 describes some of the most promising recent architectural innovations in deep learning models. It then walks you through the theory and practice of refining existing models trained by others and gives you advice on how to refine and deploy models in production:
- Time-series modelling with recurrent networks; LSTM; transformers
- Fine-tuning pre-trained PyTorch models with HuggingFace
- Real-world advice on improving ML models; diagnostics
- Deploying machine learning models in production
- Overview of the DL package ecosystem in Python; trends, innovations: Lightning; AutoPyTorch; Jax; Elegy
Personal help
We are happy to offer on-the-spot problem-solving after each day of the training for you to ask one-on-one questions — whether about the course content and exercises or about specific problems you face in your work and how to solve them. If you would like us to prepare for this in advance, you are welcome to send us background info before the course.
Other information
-
Format:
-
Courses are conducted online via live video meeting and using Python Charmers' cloud notebook server for sharing code with the trainer(s).
-
Computer:
Hardware: we recommend ≥ 8 GB of RAM and a webcam. Preferably also multiple screens and a quiet room (or headset mic).
Software: a modern browser: Chrome, Firefox, or Safari (not IE or Edge); and Zoom.
-
Timing:
Most courses will run from 9:00 to roughly 17:00 (AEST) each day, with breaks of 50 minutes for lunch and 20 minutes each for morning and afternoon tea.
Certificate of completion:
We will provide you a certificate if you complete the course and successfully answer the majority of the exercise questions.
Materials:
You will have access to all the course materials via the cloud server. We will also mail you cheat sheets and a USB stick with all the materials for reference.
Testimonials
“The course was delivered by trainers who were extremely knowledgeable in their field. It was really good to learn from the best!”
“Great course. Enjoyed it and will follow up with some practical implementation of some of the work.”
“One of the best training courses I've been on.”
“I loved it. Ed was inspiring.”
“Really impressed by Python's capability and excited to use as alternative to MatLab, as is free and better supported.”
“Very comprehensive intro to every aspect of python. Highly qualified trainer. Beyond my expectation on every aspect.”
“The VM setup and USB is great. Ed is an excellent instructor - he presents well and welcomes any questions. He is clearly a super smart guy who has a great grasp on what he is teaching - able to just prototype on the fly and the course overall really opened my eyes to python.”
“Course content was well presented and easily digested. Practical exercises were an essential part of the course – a good ratio of lecture/play was achieved. Well done Ed and Henry!”
“Simply awesome!!”
“Both Ed and Henry presented well…. The course structure was adjusted to suit the participants quickly and easily.”
“Excellent training course, excellently presented. Perhaps the best that I have had in the area of IT / programming.”
“One of the best programming courses I have attended - thanks!”
“This course has shown me how I could have done the work I was doing just last week 10x more efficiently in Python.”
“It was a pleasure ... Shared feedback from all involved was that it’s been one of the most beneficial and well delivered training courses we’ve been a part of.”
“Was the most fulfilling and rewarding class I have taken since "general relativity" at uni. Was extremely well run. Excellent all round!”
“Very impressed with the course, delivery. And depth of knowledge of Ed and Henry. Far exceeded my expectations and has greatly improved my core skills as well as inspired so many new ideas for my current work / projects. Thank you!”
