Atlas record: https://www.aitimeline.dev/en/events/evt-0254/
Availability · November 2012
Theano
The Montreal group released a compiler for mathematical expressions that computes derivatives itself and moves work onto the graphics card itself.
Why it matters
Automatic differentiation became infrastructure: a researcher described a model instead of writing gradients by hand.
Before Theano everyone derived and coded their own derivatives, and an error in the gradient was a common reason a network failed to train. Theano built a symbolic graph, differentiated it and compiled it for the available hardware. That scheme is the basis of TensorFlow. Development stopped in 2017, when more broadly supported systems appeared.
Event record
- Event date
- November 2012
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 17, 2026
- Lines
- ID
- evt-0254
The paper on new features is dated November 2012; the first Theano paper appeared at SciPy 2010.
Related events
Earlier
Later
- Related Caffe 2014links to this record
A second answer to the same need, this one aimed at vision.
- Builds on Keras 2015links to this record
The first version ran on top of Theano.
- Related TensorFlow 2015
The direct forerunner of TensorFlow in design: a symbolic graph with automatic differentiation.
- Related The first national AI strategy 2017links to this record
The same Montreal group that made Theano.