Deep Learning with the Analytical Engine

Explore the potential of Charles Babbage's Analytical Engine to perform deep learning in this 36-minute Churchill CompSci Talk. Delve into the history of mechanical computation, from the Difference Engine to the Analytical Engine, and discover how 21st-century algorithmic intelligence can be implemented on a 19th-century machine. Learn about neural networks, the MNIST classification problem, and various activation functions while following the journey of adapting modern deep learning techniques to Victorian-era technology. Gain insights into the hardware specifications, memory, and speed of early computers, and consider Ada Lovelace's prescient thoughts on machine intelligence and the Information Age. Through this immersive Victorian adventure, encounter the challenges and possibilities of combining historical mechanical computing with cutting-edge artificial intelligence concepts.

Intro
Charles Babbage
Soirée (high-level design)
Mechanical addition
The Difference Engine How would you compute the sequence of cubes?
The Analytical Engine
Difference Engine versus Analytical Engine
Timeline of mechanical computers
Hardware specifications - memory
Hardware specifications - speed
Lovelace on the Information Age
Lovelace on Machine Intelligence
Neural networks
The MNIST classification problem
Simple neural network for MNIST
What to use for the activation function?
The tanh(x) activation function
Improving the activation function
How about two convolutional layers?
Implementation
Acknowledgements