A Very Short History of Artificial Neural Networks

James V Stone
Apr 21 · 10 min read
Figure 1. A taxonomy of artificial neural networks.
Figure 2. The simplest perceptron, with two input units and one output unit. The connections from input to output units have weights w1 and w2.
Figure 3. Content addressable memory. Even a simple neural network can recover a previously learned image (right) from a corrupted version of that image (left). The image on the left was used as input, and the image on the right is the output. From Artificial Intelligence Engines, 2019.
Figure 4. A Hopfield net with seven fully interconnected binary units.
Figure 5. A Boltzmann machine with four input units, two hidden units, and four output units. This configuration is known as a 4–2–4 autoencoder.
Figure 6. Training a Boltzmann machine consists of an outer loop and an inner loop using simulated annealing.
Figure 7. A backprop neural network with two input units and one output unit. For each learned association, the error at the output is propagated back to hidden units These errors are then used to learn correct weights between units.

Reinforcement Learning

Figure 8. Reinforcement learning can be used to balance a pole on a cart by nudging it left or right. Reproduced with permission from https://github.com/david78k/pendulum.
Figure 9. Learning to soar using reinforcement learning. (a) Glide used for learning. (b) Before learning, flight is disorganised, and glider descends. c) After learning, glider ascends to 0.6 km. Note the different scales on the vertical axes of b and c. Reproduced with permission from Guilliard et al. (2018); (b,c) from Reddy et al. (2016).

From Backprop to Deep Learning

Figure 10. A deep network with three hidden layers.

References

James V Stone

Written by

James V Stone is an Honorary Associate Professor at the University of Sheffield, England. Published books: https://jim-stone.staff.shef.ac.uk

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade