Deep learning: proliferative index of a cancer cell with microscopy
M Hallett
21/07/2020
Goal today
Show a few examples of deep learning in different domains
Explore a little bit about how it works
Show an example of supervised learning using microscopy and breast cancer cell lines
Examples
These systems learn a task without being given implicit instructions. This is an unsupervised approach.
Convolution neural networks (CNNs)
There is considerable high quality material out there covering the majority of subjects related to deep learning.
I highly recmomend fastai, an excellent accessible course from UCSF that does not require sophisticated mathematis nor computing power.
Lecture 1, 23mins Jupyter (R Notebook), Python (R), PyTorch (Keras)
Lecture 2, 22-28mins, Transfer learning
Lecture 2, 1hr18mins, Data wrangling for deep image analysis
Lecture 2 1hr20mins and start of Lecture 3, Supervised learning
Lecture 5, Ethics of deep learning
There are also many mini-lectures on different topics that are quite accessible.
Here is one on Convolution Neural Networks (CNNs) which is the topic for today.
Transfer learning
Image to Image Translation.
Consider microscopy for example.
Fun examples
The following are interesting fun examples that might appear at first irrelevant to biology or science in general. However that is not likely the case. The basic techniques underlying each could be applied in a braod spectrum of life science domains.
Some examples in biology
Several movies/documentaries of interest
BIOL 480
© M Hallett, 2020 Concordia University
