Questions tagged [topology]

For questions involving topology in any form in relation to Artificial Intelligence.

Topology is central to several fields of AI application.

  • In solid modelling, ray tracing, dynamic morphing, mechanical design, 3D printing, and numerically controlled (CNC) machining, topological analysis of surface adjacency and cornering is necessary to ensure reliable control.

  • In networking, the topology of a scalable network is often more important than the network's scale when projecting network behavior for capacity planning.

  • In machine learning designs, the interconnection of networks, such as feeding a CNN network used to recognize object movement into an RNN or LSTM, guiding Q-learning using a trained MLP, or using generative designs that involve oversight or balance between multiple networks all form a topology with the surroundings that is the most fundamental aspect to high level design.

  • In drawing electronic schematics or laying out printed circuit boards or integrated circuits, the topology of the circuit determines the ease and potential space and path efficiency of the layout. CAD software represents the circuit and its functioning as a directed or non-directed graph. Electronic and electrical CAD automation involves translating elementary components in a multidimensional space into two dimensions of a fixed set of circuitry and component layers.

  • In natural language recognition and authoring, the serial stream of sound or printed characters must be probabilistically parsed into a semantic network of objects and associations and the semantic network must be serialized into sound or printed characters. This is central to social AI, automation of call centers and chatbots, and the development of conversant robots.

  • Ontologies have a topology that is related to complexity and the features of line topology define the complexity of object design that represents the ontology and related operations.

  • The brains of people and animals contain topological structure in the division of brain regions, in the wiring of neurons, in the channels of chemical signaling, and the plastic structure of axons, dendrites, and synapses that are related to function and timing.

The Oxford Dictionary provides this definition:

to·pol·o·gy
/təˈpäləjē/
noun
1. the study of geometric properties and spatial relations unaffected by the continuous change of shape or size of figures.
2. the way in which constituent parts are interrelated or arranged: "the topology of a computer network"

The popularity of arrays, vectors, lists, loops, and tensors is a topological phenomenon in that computer machinery and language has evolved away from relay racks containing a rat's nest of relay wiring toward structures that are more rectilinear and easier to comprehend using tools of mathematics and language designed for orthogonal structures. The popular programming languages tend away from symbolic processing languages that handle arbitrary topology like LISP to languages that are heavily influenced by FORTRAN (formula translation to assembly language) with arrays and loops as the central parallel structure methodology.

Graph theory and graph software libraries support the representation of non-orthogonal structures, such as clusters of highly interrelated neurons found in brains that don't fit nicely into traditional tensor representation. A calculus of topological structure may or may not facilitate further and faster advancement of AI capabilities. Projections are divided and proposed process topology and data flow topology is varied.

In AI theory, many have proposed changes in computing topology away from the approach initiated by John von Neumann and others, which centralizes arithmetic and logical branch processing in a CPU. The argument points out that both processes and processors have topology and that a close alignment of the two produces greater efficiency in parallel processing in VLSI design, mother board design, and computer clustering. If this principle is generally true, energy efficiency, speed of computation, reliability, weight (for launch application), and conservation of space are all related to topology.

16 questions
19
votes
2 answers

Are Modular Neural Networks more effective than large, monolithic networks at any tasks?

Modular/Multiple Neural networks (MNNs) revolve around training smaller, independent networks that can feed into each other or another higher network. In principle, the hierarchical organization could allow us to make sense of more complex problem…
11
votes
1 answer

Can layers of deep neural networks be seen as Hopfield networks?

Hopfield networks are able to store a vector and retrieve it starting from a noisy version of it. They do so setting weights in order to minimize the energy function when all neurons are set equal to the vector values, and retrieve the vector using…
8
votes
2 answers

In what ways is the term "topology" applied to Artificial Intelligence?

I have only a general understanding of General Topology, and want to understand the scope of the term "topology" in relation to the field of Artificial Intelligence. In what ways are topological structure and analysis applied in Artificial…
DukeZhou
  • 6,237
  • 5
  • 25
  • 53
4
votes
1 answer

Neural networks of arbitrary/general topology?

Usually neural networks consist from layers, but is there research effort that tries to investigate more general topologies for connections among neurals, e.g. arbitrary directed acyclic graphs (DAGs). I guess there can be 3 answers to my…
TomR
  • 823
  • 5
  • 15
4
votes
1 answer

What topologies support recognition of action sequences?

The ability to recognize an object with particular identifying features from single or multiple camera shoots with the temporal dimension digitized as frames has been shown. The proof is that the movie industry does face replacement to reduce…
2
votes
0 answers

zonal or template ocr invoices reading

I'd like to explore the possibilities of applying artificial intelligence to ocr reading. Basic ocr invoices processing let me convert 30% of them only. The main purpose is defining invoices areas by training an ai, then process those areas with…
2
votes
1 answer

Is a calculus or ML approach to varying learning rate as a function of loss and epoch been investigated?

Many have examined the idea of modifying learning rate at discrete times during the training of an artificial network using conventional back propagation. The goals of such work have been a balance of the goals of artificial network training in…
2
votes
0 answers

Nearest neighbour search in high dimension retrieves certain points too often

We represent some catalogue items (documents, music tracks, videos, whatever) as vectors embedded in R^d and use them to retrieve nearest neighbours to users query. The typical scenario is that users can input any query and the search results are…
1
vote
1 answer

How can AI be used to more reliably analyze and plan around the tie between climate and emissions?

Note to the Duplicate Police This question is not a duplicate of the Q&A thread referenced in the close request. The only text even remotely related in that other thread is the brief mention of climate change in the Q and two sentences in the sole…
Douglas Daseeco
  • 7,423
  • 1
  • 26
  • 62
1
vote
0 answers

Comparing Auto-regressive Encoder-Decoders and Topological Neural Networks

I am interested in what insights can be gained about the mathematical class of auto-regressive encoder-decoders (LLMs), by comparing them to topological neural networks. Specifically, I am looking for similarities and differences in their…
hmltn
  • 103
  • 9
1
vote
0 answers

How do I use the N correctly in NEATs speciation delta function?

When implementing NEAT I'm having some issues with the speciation distance/delta function, specifically the term N (number of genes in biggest genome). Won't term $N$ in $δ=c1*E/N+c2*D/N+c3*W$ just out-scale $E$ and $D$, and reduce their impact on…
1
vote
1 answer

Are there any animation tools available to visualise and simulate deep neural networks?

Deep learning researchers have to work with a lot of models. The models may include different types of Layers: They include convolutional neural network layers, recurrent neural network layers, batch normalization layers, polling layers, and many…
1
vote
0 answers

Is there any application of topology to deep learning?

Is there any application of topology (as in math discipline) to deep learning? If so, what are some examples?
SpiderRico
  • 960
  • 8
  • 18
1
vote
0 answers

Number of input variables for a cellular automaton (was: Squares or hexagonal?)

A cellular automaton is a state machine that is controlled by external input. The input is given by geometrical space around a cell. In a square matrix, each automaton gets input from 4 surrounding cells, while a hexagon grid has 6 neighbor cells…
kenorb
  • 10,423
  • 3
  • 43
  • 91
0
votes
0 answers

Behaviour of PPO/similar Algos under action penalties

I am currently experimenting with PPO in different environments. I am interested in learning policies that fulfill a certain goal while keeping a specific value low. Here's an example: Using PPO on a cartpole environment to learn an upswing of the…
1
2