The master branch is ported to Java 10, and uses specificities of it so can't be compiled nor ran on jdk < 10. The JAVA8 branch can be compiled and ran on jdk 8/9/10 (there are maven "auto-profiles" for those).

Java port of the "Simulator for digital circuits" from SICP

A Java port of the "Simulator for digital circuits" from SICP (Structure and Interpretation of Computer Programs). The full book is available online: https://mitpress.mit.edu/sicp/full-text/book/book.html I have the paper edition, but it's nice to be able to link to a non-paywalled, full edition, so thanks MIT for that!

The idea of this project comes from a friend of mine mentioning connecting neural networks together. It reminded me of doing it like in that chapter (rather than using Kafka or any complex frameworks): https://mitpress.mit.edu/sicp/full-text/book/book-Z-H-22.html#%_sec_3.3.4 Even though Lisp and Java obviously are very different, with just Java 8 lambdas it's not hard to implement almost exactly the same concepts. So then, it shouldn't be that hard to implement a neural network from that, right??? (yeah right) For now (and for once) I wanna keep from looking at how frameworks like DL4J/TensorFlow/whatever do it. Instead I'm going to try to use: -SICP concepts (wires, ...) as low-level bricks -to implement the model from the French book "Réseaux neuronaux" by Jean-Philippe Rennard (Editions Vuibert) -using my Java experience to do it in a more object-oriented way than in the book

See:

• Classes in /src/main/java/fr/cla/wires/boxes/exampleusage/ for example usages of how to connect Wires to Boxes
• Tests in /src/test/java/fr/cla/wires/boxes/exampleusage/ for expected behaviour and examples of how to tick the clock.
  • Tests in /src/test/java/fr/cla/wires/boxes/exampleusage/pbt/ use Property-Based Testing to assert the invariants of various Boxes.

(Developper note)

If you're wondering what the //@formatter:off //@formatter:on comments are all about, it's because:

• Even though IntelliJ is overall a better IDE than Eclipse IMO, it's also more annoying in some regards (La perfection n'est pas de ce monde),
  • generally in that it sometimes tends to be too opinionated,
  • and specifically in that unlike Eclipse, it doesn't have a global, Don't reformat on save / Thanks, but I know what I'm doing option (at least give us options: "Never" OR "Never for Java files", since this issue is more pregnant in the "core").
    • and they outright say they won't support it.
• So instead of just configuring IJ, I have to sprinkle it all over the place. Overall thanks IJ for the stable IDE experience, but please be less apple-like.