-
Notifications
You must be signed in to change notification settings - Fork 0
Neuromorphic Computing
<todo: tidy up. a lot>
Nanoscale Brain inspired Artificial Synapses 2022 analog deep learning
Neuromorphic Computing, NEURON software
Brain Inspired Highly Scalable Neuromorphic Hardware, cointegration of single-transistor neurons and sysnapses with standard CMOS semiconductors
DeepDendrite
Memory boosting
Biophysical - Flatrion Institute for Computational Neuroscience
Linear systems - environment, biochemical signalling,
Robustness - quality, ignore extringic noise,
Computatonal Memory Models
Statistical models
Text mining as a case in point. <todo: tidy up content in this section.>
The problem of unstructured data and no common ontology/taxonomy/data dictionary standards persists.
As does the issue of coining of the new terms in many studies. As a way for the project practitioners to make sense of a somewhat opaque domain specific information.
The issue of what scales of space and time in relation to thing studied continues. Studying a part of the brain at different levels of scale of matter and therefore corresponding time in i/o and interactions. From whole body inter system interactions with the endocrine system or immune system for example. To external environmental interactions like for example with the electromagnetic spectrum like mobile phone masts (cell towers, base stations) to electricity transmission cables to microwaves from mobile phones themselves. To chemical elements and valance bonds and molecular orbitals at the quantum chemistry level.
Which might go to wider concerns of policy issues re production of or quality of or contamination of food and water and so on. And the gastrointestinal system and metabolic pathways from farm to food to fork to absorption into the whole body system and transmission to the brain. Say for example mercury Hg in fish. Or lead Pb particulates in the air. Or metals in food from cookware like iron Fe or aluminium Al. Or contamination of the water H2O supply by some means possibly bacterial like legionnaires’ disease and neurological impacts like encephalopathy or seizures or loss of coordination and so on due to brain lesions.
So while while many studies might address the same brain region the focus of the study is likely as or more important to consider.
Yes. Depends on the objective of the text mining exercise. That is what is text mining activity for?
Macro. A board wide assessment of the entire literature can be useful for creating a large sample set for the entire neurological field. It would also serve to show ontological wormhole links to related domains. It might be used also as a means to assess compliance with standards in nomenclature or design of experiments and so on. Quality issues.
Micro. A narrow deep assessment of domain specific sub set of the literature might be useful for the first steps in creating a domain specialist specific ontologies. Or finding best practice or new practice or bad practice in a sub field of study of neurology.
Meso. A mid range assessment of the literature across document management or record management of different types. For metadata or reference data. Or for technology specific things like data formats and ability to manage structured and semi-structured and unstructured data.
It will largely depend on the purpose of the text mining initiative.
But as a general heuristic; don’t try and boil the ocean. For a general principle start with a defined project/business objective. Fit the information gathering to the objective. Start with a core set of terms say twenty and work outward. Beware scope creep and mission creep. So determine success criteria up front, hypothesis to prove, re falsifiability. Iterative and incremental deliverables to review meeting of project objectives.
Good project and programme management and architecture governance should help keep things on track. Ensure backlog for issues/risk and use of wikis and so on for genesis of next state research topics as a consequence of discovery on current state topic.
Subject project wikis and project documentation outputs to similar text mining for IT governance and compliance and continuos improvement. Meta architecture of meta studies. How to improve the state of the art for neuroinformatics, other related science disciplines, state of the art in general, and so on. But this wider effort is also a funding issue so depending of funding project pipelines may be considered second order by an organisation. Which would be a missed opportunity.
System of system SoS viewpoints. Start with the meta architecture standard;
Systems and software engineering — Architecture description ISO/IEC/IEEE 42010 http://www.iso-architecture.org/42010/
This standard is a foundational part of both TOGAF and UAF. So on solid ground here.
Various grids of fields of specialism within the domain of neuroscience. Various grids of tools/api’s/libs/applications/ hardware software.
These to provide an enterprise architecture view of specialism to tool set. And wider neuroscience technology stacks both hardware an software related. So akin to LAMP software bundle; Linux, Apache, MySQL, Perl/PHP/Python . But inclusive of hardware like SpiNNaker and wafer technology and so on. And cloud based XaaS offerings. And information space standards like ontologies/taxonomies and so on.
From a biological SoS viewpoints of whole body brain interactions. Body sub systems brain interactions like immune system. Brain and brain region interactions. And so in a reductionist manner with the creation of several holarchies and hetearchies. So environmental to quantum chemical. Related maps to fields of expertise and disciplines in neuroscience.
As well as technology SoS viewpoints of technology stacks for each of the biological SoS viewpoints. Some of which would be in relation to text mining the literature.
Further a FINOS Linux Foundation type initiative for open source projects in the neuroinformatics domain.
Support for foreign languages is a language translation issue. It is not a core text mining issue. Although they are related in regards to data governance and origination and ownership and so on. Also to natural language processing NLP applications. And related issues.
Translation is an important topic of concern. To keep it simple and don’t repeat yourself are two important IT principles. The language translation and text mining topics of concern should be kept quite separate.
The data management pipeline inclusive text mining and text language translation be different activities in the pipeline process.
Syntax (morphemes, word order, phrases, sentences, grammatical rules, … ) and semantics (meaning) are different things to consider. The meaning part must be objectively agreed. Which is why common ontology standards and the like are so important.
The provenance and quality of source material, data, as other respondents have discussed is a core consideration. Otherwise garbage in garbage out.
One might have the best data pipeline ever possible in all time but if the inputs are of low quality so will the outputs be low quality.
And quality control is very important to maintain credibility. A lack of quality control and poor or bad outputs will eventually lead to poor and bad medical practice. In which case neuroscience will be in danger of being discredited and suffer gross reputational risk due to bad down stream medical practice or malpractice. It is a train of trust topic of concern. This is in part a complexity issue and also an ethics (research ethics and medical ethics) issue and regulatory issue.
The greater the complexity the more likelihood of entropy affecting the system at some point. And the entropy not being spotted down stream due to high degrees of specialism at each point of activity. An information corruption cascade affect.
Relying on straight through processing STP with ChatGPT with no oversight or validation would be highly likely to lead to information corruption. STP with ChatGPT would likely be a guaranteed recipe for failure.
The erosion of trust in science and the scientific community must be robustly tackled. Successfully overcoming the difficulties of global warming alone require it.
This could quickly become a Ouroboros problem. The snake eating its own tail. Which it already has been in relation to generative AI and misinformation disinformation on social media. Profit before responsibility.
How many papers on self publishing sites have been created with the aid of Big Tech NLP for example. Where all the sources cited are of unclear quality and provenance. But still fool journals.
Science is already facing reputational challenge from many quarters which seek to discredit it.
It is not beyond the bounds of reason to suppose that some bad actors might want to use Big Tech NLP for bad purposes. Like discrediting science by flooding science journals with ‘credible’ looking ‘paper mill’ papers.
This was exposed years ago by Alan Sokal in the infamous Sokol affair in 1996. And many others besides. See also includes many other similar hoax papers and problems. https://en.wikipedia.org/wiki/Sokal_affair
The postmodern generator was a programme created for making academese.
Which all leads to the demarcation problem. What is and is not science.
NEWS FEATURE 23 March 2021 The fight against fake-paper factories that churn out sham science Some publishers say they are battling industrialized cheating. A Nature analysis examines the ‘paper mill’ problem — and how editors are trying to cope. By Holly Else & Richard Van Noorden https://www.nature.com/articles/d41586-021-00733-5
Body system of systems SoS viewpoints, Body whole/part viewpoints, body environment SoS view, body system brain axis, body organ part of a body system brain axis. Holarchies heterarchies.
From a whole body view all organs, inclusive brain, and all systms, inclusive cns/pns/ens, are parts of the whole body. Which organ malfunctions cause brain response malfunction or injury? What is the order of precedence of other organ/system malfunction/disease and affects on brain cns/pns/ens?
- Heart brain axis, cardio vascular system
- Gut brain axis, gustatory system, ens,
- Lung brain axis, pulmonary/respiratory system
- Lymph node brain axis, adaptive immune system
- Hormone brain axis, endochrine system, pineal gland, pituitary gland, thyroid gland, thymus, pancreas, adrenal gland, gonad (ovary, testis),
- Vagus nerve, <todo: group otherwise?>
- others to list
<todo: look into following if true look interesting, test validity statements, response to post in Simulation Neuroscience, Notch, MU_2410_F3YA>, WS, heart brain axis, gut brain axis, gut bacteria producing neurotransmitters such as serotonin and GABA, effects on mood and behaviour, vagus nerve, signals related to digestion and emotional states, mental health considiton linked to gut health, axiety, depression, neurodegenerative diseases, cognitive function affected by inflammatory responses in brain can be response to inflammation in the gut, psychological activities occuring in brain directly related to digestion, data simulation for curing degenerative brain disease that impact whole body,
!! find tool chain, api chain, software chain, ontologies, hierarchies space/time, i.e. LAMP(,s) for neurosience, does a grid exist?
Ontology - schema, taxonomies, data dictionaries
- Minimal Information for Neuroscience Data Standard MINDS, GH, WS schema, INCF,
- Open Archives Initiative WS,
- <todo: get other realted things from in silico>
Database
- CKAN WS data management system, open source, open knowledge foundation
Software - anatomy
- Hippocampome.org ?
Software - blue brain project
- Blue Brain Project GH
- Bluima, GH, NLP for neuroscience, Blue Brain project, document parsing, science papers,
- NeuroCurator, literature anotation, code data values from literature, Blue Brain project, <todo: find it, was it deleted? renamed? refactored? >
Hardware
- Field Programmable Gate Arrays FPGA
- Application Specific Integrated Circuits ASIC
- FACETS Flexible Analog Circuit Technology for Emulation and Simulation, wafer-scale hardware system, analog neuron and synapse circuits, builds on BrainScaleS,
- ANC Analog Network Chip
Differential computing
- SpiNNaker project WP, digital circuit and distributed architecture, Human Brain Porject,
- SpiNNaker 1, WS, PyNN API, Manchester University, SNN's
- SpiNNaker 2, WS, GL aka BIMPA, PyNN API, NEST, University of Dresden, SNN's, DNN's,
- SpiNNcloud, SpiNNaker 2 base, ARM chip, Globally Asynchronous Locally Synchronous GALS,
- SpiNNcloud Systems, WS
- SpiNNcloud Platform, based on SpiNNaker 2 (BIMPA), ~circa 10 million neurons simulation,
Lib - neural networks, circuits, layer columns, <todo: what is LAMP equivalent tool sets for neurosience. >
- PyNN, 'pine', GH, lib, Python, various NN types, multiple simulator support, neuromorphic hardware system support, connection-set algebra,
- NineML, lib, connection-set algebra,
- Connection-Set Algebra, CSA, Python, INCF GH, lib,
- Connection-Set Algebra, CSA, Haskell, WS, lib,
Lib - anatomy, imaging <todo: relation to PyNN or similar? use by PyNN?>
- Voxel Brain API, for brain atlas creation, and neural circuits, Python, data formats; NRRD, JSON, CSV, . Algo; DARTEL, VBM
- FiftyOne Brain, large scale brain dataset analysis
Simulator
- NEST, C++, SNN, simulator, stand alone, command line interface,
- NEURON, simulator
- Brain, simulator
- Arbor, simulator
Org
- Neural Ensemble org, WS, open source, neurosicence, home of PyNN
- International Neuroinformatics Coordinating Facility org, INCF, GH, WS,
Gov
- Human Brain Project HBP
- EBRAINS (European Brain Research Arm for Innovation, Networking, and Science), WS, successor to HBP
Edu
- Neurogrid - Stanford
- BrainScaleS, wafer scale neuromorphic hardware system, University of Heidelberg, Technische Universität Dresden
- SpiNNaker - Manchseter, Heidelberg, Dresden, HBP, EBRAINS,
- Neomorphic computing, WP
- SMIRN, memory model?
- Tensor network theory
Datamining
- Extractors
- Named entity recognition NER,
- Connectivity satement, formalisation? relation to SPO in RDF?
Mathematics