Piotr Migdal: "Learning neural networks within Jupyter Notebook"
Learning neural networks within Jupyter Notebook
Are artificial neural networks hard, only for a handful of people possessing some forbidden knowledge?
Well, either there aren't, or it is possible to learn some of these hidden secrets in 45 min!
I will show you that, creating a convolutional neural network to distinguish drawings of lightbulbs, candles and campfires.
I will use Keras, a popular high-level deep learning framework, in Jupyter Notebook, an interactive Python environment awesome for learning.
I will sprinkle it with a number of libraries & scripts, that make it easier, more interactive and visual.
Including livelossplot - my Python package for live charts that track the learning process, and ASCII art for visualizing network architecture.
Piotr Migdal
Польша. Варшава
Data science consultant
p.migdal.pl
Piotr is an independent data science consultant focusing on machine learning, deep learning and data visualization. He holds a PhD in quantum physics from ICFO, Barcelona.
Piotr worked on numerous projects with deepsense.ai on computer vision and gave trainings to companies such as Intel and BCG. He collaborates with RaRe Technologies on natural language processing workshops. Piotr is the author of a popular blog post series introducing readers to data science:
Data science intro for math/phys background
king - man + woman is queen; but why?
Learning Deep Learning with Keras
Starting deep learning hands-on: image classification on CIFAR-10
In his free time, he created the Quantum Game with Photons and is a volunteer teacher of gifted high-school students. He founded the Data Science PL Facebook group - the biggest such community in Poland. Piotr has lectured at Imperial College London and given talks at Caltech and the Bay Area D3.js User Group, among other places.
He authored open source Python packages:
keras_sequential_ascii - sequential model in Keras -> ASCII diagrams
livelossplot - live training loss plot in Jupyter Notebook for Keras, PyTorch and other frameworks
Blog GitHub Twitter