Lambda calculus inspires experiments with chemlambda

In the now deleted chemlambda collection I told several stories about how lambda calculus can bring inspiration for experiments with chemlambda. I select for this post a sequence of such experiments. For previous related posts here see this tag and this post.

Let’s go directly to the visuals.

Already in chemlambda v1 I remarked the interesting behaviour of the graph (or molecule) which is obtained from the lambda term of the predecessor applied to a Church number.  With the deterministic greedy algorithm of reductions, after the first stages of reduction there is a repeating pattern of  reduction, (almost) up to the end. The predecessor applied to the Church number molecule looks almost like a closed loop made of pairs A-FO (because that’s how a Church number appears in chemlambda), except a small region which contains the graph of the predecessor, or what it becomes after few rewrites.

In chemlambda v2 we have two kinds of fanouts: FO and FOE.  The end result of the reduction of the same molecule, under the same algorithm, is different: where in chemlambda v1 we had FO nodes (at the end of the reduction), now we have FOE nodes. Other wise there’s the same phenomenon.

Here is it, with black and white visuals

pred

Made by recording of this live (js) demo.

1. What happens if we start not from the initial graph, but from the graph after a short number of rewrites? We have just to cut the “out” root of the initial graph, and some nodes from it’s neighbourhood and glue back, so that we obtain a repeating pattern walking on a circular train track.

Here is it, this time with the random reduction algorithm:

bigpred_train-opt

I previously called this graph an “ouroboros”. Or a walker.

2. That is interesting, it looks like a creature (can keep it’s “presence”) which walks in a single direction in a 1-dimensional world.  What could be the mechanism?

Penrose comes to mind, so in the next animation I also use a short demonstration from a movie by Penrose.

bigpred_penrose-opt

 

3. Let’s go back to the lambda calculus side and recall that the algorithm for the translation of a lambda term to a chemlambda molecule is the same as the one from GLC, i.e the one from Section 3 here. There is a freedom in this algorithm, namely that trees of FO nodes can be rewired as we wish. From one side this is normal for GLC and chemlambda v1,  which have the CO-COMM and CO-ASSOC rewrites

convention_3

In chemlambda v2 we don’t have these rewrites at all, which means that in principle two diferent molecules,  obtained from the same lambda term, which differ only by the rewiring of the FO nodes may reduce differently.

In our case it would be interesting to see if the same is true for the FOE nodes as well. For example, remark that the closed loop, excepting the walker, is made by a tree of FOE nodes, a very simple one. What happens if we perturb this tree, say by permuting some of the leaves of the tree, i.e. by rewiring the connections between FOE and A nodes.

bigpred_train_perm-opt

The “creature” survives and now it walks in a world which is no longer 1 dimensional.

Let’s play more: two permutations, this time let’s not glue the ends of the loop:

bigpred_train_egg

It looks like a signal transduction from the first glob to the second. Can we make it more visible, say by making invisible the old nodes and visible the new ones? Also let’s fade the links by making them very large and almost transparent.

bigpred_train_egg_mist_blue

Signal transduction! (recall that we don’t have a proof that indeed two molecules from the same lambda term, but with rewired FO trees reduce to the same molecule, actually this is false! and true only for a class of lambda terms. The math of this is both fascinating and somehow useless, unless we either use chemlambda in practice or we build chemlambda-like molecular computers.)

4.  Another way to rewire the tree of FOE nodes is to transform it into another tree with the same leaves.

bigpred_tree-opt

 

5. Wait, if we understand how exactly this works, then we realize that we don’t really need this topology, it should also work for topologies like generalized Petersen graphs, for example for a dodecahedron.

dodecahedron_walker

 

This is a walker creature which walks in a dodecaheral “world”.

6. Can the creature eat? If we put something on it’s track, see if it eats it and if it modifies the track, while keeping it’s shape.

walker_bit-opt

So the creature seems to have a metabolism.

We can use this for remodeling the world of the creature. Look what happens after many passes of the creature:

walker_bit_new

 

7. What if we combine the “worlds” of two creatures, identical otherwise. Will they survive the encounter, will they interact or will they pass one through the other like solitons?

bigpred_bif

 

Well, they survive. Why?

8. What happens if we shorten the track of the walker, as much as possible? We obtain a graph wit the following property: after one (or a finite give number of) step of the greedy deterministic algorithm we obtain an isomorphic graph. A quine! chemlambda quine.

At first, it looks that we obtained a 28 nodes quine. After some analysis we see that we can reduce this quine to a 20 nodes quine. A 20-quine.

Here is the first observation of the 20-quine under the random algorithm

20_quine_50steps

According to this train of thoughts, a chemlambda quine is a graph which has a periodic evolution under the greedy deterministic algorithm, with the list of priority of rewrites set to DIST rewrites (which add nodes)  with greater priority than beta and FI-FOE rewrites (which subtract ndoes), and which does not have termination nodes (because it leads to some trivial quines).

These quines are interesting under the random reduction algorithm, which transform them into mortal living creatures with a metabolism.

____________

So this is an example of how lambda calculus can inspire chemlambda experiments, as well as interesting mathematical questions.

One thought on “Lambda calculus inspires experiments with chemlambda”

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s