Multimodality - Science topic

How can we address the unique challenges of managing pain in older adults?

How can interprofessional communication improve patient outcomes?

What are the long-term risks associated with chronic opioid therapy?

How can patient adherence to nonpharmacological therapies be improved?

What are the current best practices for minimizing opioid use in chronic pain?

How can healthcare teams collaborate to provide comprehensive pain care?

What are the most effective methods for diagnosing chronic pain?

Personalized and real-time image captioning enhances user experience by adapting captions to preferences and delivering dynamic descriptions for changing content. Personalized systems leverage user profiles, fine-tune models on specific data, and incorporate feedback loops or natural language understanding for tailored outputs, benefiting accessibility tools, e-commerce, and social media. Real-time captioning uses low-latency models, temporal analysis, event detection, and multimodal inputs to generate fast, accurate captions for videos, live streams, and dynamic environments like surveillance or education. While challenges like privacy, scalability, and latency persist, advancements in ethical AI and optimized architectures promise seamless and user-centric solutions.

ChatGPT reveals that while its story begins around 2015, its current capabilities are the result of years of research, development, and most significantly learning from vast amounts of data, per the below.

· GPT-1 – June 2018 (117 million parameters)

· GPT-2 – February 2019 (1.5 billion parameters)

· GPT-3 – June 2020 (175 billion parameters)

· GPT-3.5 – November 2022 (further refinements on GPT-3)

· GPT-4 – March 2023 (multimodal, improved reasoning)

· GPT-4 Turbo – November 2023 (faster, more cost-efficient variant)

Turbo, the last version, is the prime engine processing all queries since its release, both paid and unpaid. This vast amount of data includes the near totality of human savoir-faire, professional and scientific knowledge bases in all fields, to the point that it can pass strict professional exams and write theses at the doctorate level.

The question is: with this humongous amount of data, and their extensive language-based reasoning capabilities, why have we not seen any scientific breakthrough by these LLM’s in nearly 15 years of fending on their part altogether? Does that say something about our model of science (scientific method), and the value and validity of what we know in science, in particular the fundamental premises in all disciplines? Is this a verdict on the quality of what we know in terms of our scientific principles? In light of this null result, can we expect what we know to tell us something in the least amount about or toward the resolution of what we don’t know? If there is a hard breaking between our knowns and the unknowns, can the LLM’s help at all leapfrog the barrier? Given ceiling being currently hit in their learning capacity, would more time make any difference?

I am working on a study that involves collecting synchronized EEG and eye-tracking data integrated within the iMotions software to examine cognitive workload. I have set event markers to ensure precise synchronization between the data streams. However, I’ve encountered an issue where one data stream (e.g., eye tracking) contains missing values while the other (e.g., EEG) is complete, leading to partially incomplete rows in my dataset.

I would appreciate advice on:

Any insights from those experienced with multimodal data analysis, would be extremely helpful. Thank you!

where I can get multimodal mental health dataset.

The spot diameter of the signal light is 1.3mm, and the collimating lens is a flat convex lens with a focal length of 11mm. The spot is measured by a spot quality analyzer, and the spot is measured 10cm away from the collimator. It is found that the spot is concentric circle type and the spot diameter is about 15μm after the collimating lens is focused. The large-mode optical fiber is liekki passive 30/250 dc pm with core NA of 0.07. Cut both ends of the optical fiber by 8°.

In the experiment, the collimated light beam is deflected through the half-wave plate and PBS, and then the optical fiber axis is used after passing through the second half-wave plate. The passive fiber is placed on the five-dimensional adjusting frame, and the output end of the passive fiber is collimated through a flat-convex lens with a focal length of 15mm. The output light spot can obviously see the panda eye spot, and there is no obvious bright spot in the center of the light spot, indicating that most of the signal light has entered the cladding, and the extinction ratio is only 1dB.

I have three questions:

1. If the signal is a fundamental mode Gaussian beam (actually a concentric ring type), it can be fully coupled into the core according to the formula, but the coupling effect is very poor at present, why?

2. How to measure the result of coupling? Is the output light spot periphery has entered the envelope of the diaphragm filter, and then measure the coupling efficiency, so that the extinction ratio is not taken into account. The goal of coupling is to get as much signal light into the core as possible with high coupling efficiency while maintaining a high extinction ratio for the output. The current experiments are sometimes more efficient, but the extinction ratio is worse.

3. If the passive fiber is replaced by a gain fiber, model liekki y1200 30/250 dc pm, how should the coupling result be measured? At this time, the signal light will have higher absorption in the core and less absorption in the cladding, and the coupling efficiency seems to be inappropriate.

I hope you can answer. If there are skills and experiences about spatial optical coupling into large mode field polarization-maintaining fibers, I also hope to share them.

I need Multimodal mental health data for research. I facing problem to find dataset.

Subject: Exclusive Opportunity to License or Acquire Breakthrough DIKWP-Enhanced AI Patents

I hope this message finds you well. I am writing to present a unique opportunity that could significantly enhance the capabilities of your language models, such as GPT4, by integrating advanced innovations protected by a portfolio of 90 patented technologies. These patents, developed by Professor Yucong Duan and his team, encompass cutting-edge methodologies that enhance Large Language Models (LLMs) by integrating a comprehensive DIKWP (Data, Information, Knowledge, Wisdom, Purpose) framework.

Why DIKWP Matters for your Company

The evolution of LLMs like GPT-4 has set new benchmarks in natural language processing. However, the current models face limitations in understanding complex contexts, generating goal-oriented outputs, and effectively integrating multimodal data. This is where the DIKWP-enhanced patents can revolutionize the field. The patented technologies offer:

Strategic Fit with your Company’s Vision

Your Company has consistently been at the forefront of AI innovation, and acquiring or licensing these patents would further cement your leadership in the industry. By integrating DIKWP-enhanced technologies, Your Company's LLM could:

Next Steps

I would be pleased to discuss how these patents can be integrated into your existing models and the potential for a strategic partnership. Please let me know a convenient time for a meeting or a call to discuss this opportunity in more detail.

Thank you for considering this transformative opportunity. I look forward to the possibility of working together to push the boundaries of what LLMs can achieve.

Warm regards,

Yucong Duan

Title：Evolution and Prospects of Foundation Models: From Large Language Models to Large Multimodal Models

Journal：Computers, Materials & Continua (SCI IF2.0 CITESCORE5.3)

Link：https://www.techscience.com/cmc/v80n2/57626

Since the 1950s, when the Turing Test was introduced, there has been notable progress in machine language intelligence. Language modeling, crucial for AI development, has evolved from statistical to neural models over the last two decades. Recently, transformer-based Pre-trained Language Models (PLM) have excelled in Natural Language Processing (NLP) tasks by leveraging large-scale training corpora. Increasing the scale of these models enhances performance significantly, introducing abilities like context learning that smaller models lack. The advancement in Large Language Models, exemplified by the development of ChatGPT, has made significant impacts both academically and industrially, capturing widespread societal interest. This survey provides an overview of the development and prospects from Large Language Models (LLM) to Large Multimodal Models (LMM). It first discusses the contributions and technological advancements of LLMs in the field of natural language processing, especially in text generation and language understanding. Then, it turns to the discussion of LMMs, which integrates various data modalities such as text, images, and sound, demonstrating advanced capabilities in understanding and generating cross-modal content, paving new pathways for the adaptability and flexibility of AI systems. Finally, the survey highlights the prospects of LMMs in terms of technological development and application potential, while also pointing out challenges in data integration, cross-modal understanding accuracy, providing a comprehensive perspective on the latest developments in this field.

Hi there, I'm new to modelling with comsol. I would like to ask if it is possible for comsol to output linearly polarized (LP) modes? I tried modelling a simple single mode fiber with an enlarged core so that it becomes multimode, but it seems like I can only get the exact modes (TE, TM and HE) individually for the higher order modes. Am I missing some other settings? Thanks!

how do integrate ECG, PCG, and clinical data to apply early fusion multimodal?

How does an amphibious robot with two sets of power units rationally switch between them? Or the modal switching problem for multimodal robots?

How does multimodal monitoring contribute to TBI management, and what modalities are typically included?

Discuss the use of multimodal analgesia techniques, including oral analgesics, regional techniques, and non-pharmacological interventions, for postpartum pain management.

Multimodal analgesia involves the use of multiple analgesic modalities in combination to optimize pain relief while minimizing the adverse effects associated with any single agent. In the postpartum period, multimodal analgesia is particularly beneficial for managing pain following both vaginal delivery and cesarean section.

The management of perioperative pain in paediatric patients is a critical aspect of their care, aiming to minimize discomfort, improve recovery, and reduce the risk of complications. An effective approach often involves a combination of pharmacological agents, regional anaesthesia techniques, and multimodal analgesia strategies tailored to the individual patient and surgical procedure.

Based on the my personal Gemini Ultra test results, I can say that GPT-4v is definitely better than Gemini Ultra!!!

The hype around the absolute benefits of Gemini Ultra is just a purely business PR campaign that mainly misleads users and tries to pass off wishful thinking. The multimodal capabilities of the Gemini Ultra v 1.0 are actually very limited and do not meet the requirements. At the same time, ideally, it is necessary to use these different LLMs, supplementing the gaps of one with the advantages of the other.

Please share your experience regarding this.

How can we demonstrate the efficacy of multimodal composing in enhancing writing skills, particularly in the context of academic writing, given the prevalent skepticism surrounding its effectiveness in improving academic writing skills? are there any methods or techniques of analyzing students' multimodal products to showcase improvement in terms of macro or micro-skills of writing? concerning coherence, content, organization, etc.

Hi guys,

I have a Holmium doped multimode fibre with a double cladding structure. But my available pump is only modest so without core pumping we won’t get amplification. What is the best way to check if the pump beam is efficiently coupled into the core?

The pump beam was conjugated from the 6 um core diameter SMF, expanded to about 35 um which is lightly smaller than the fibre core (40 um) using a pair of lens.

I am trying to look at the image at the other end, hoping to see the ASE glowing when the pump beam is strongly coupled to the core, as well as a minumum in total pump transmitted power. But do you know some more elegant way of doing this?

Thank you,

L

Any help

What is multimodal research?

Define what is a multimodal and what is the difference between unimodal. and how it is used in the strategy of teaching specifically in teaching science in junior high

What is the state of the art in multimodal 3D rigid registration of medical images with Deep Learning?

I have a 3d multimodal medical image dataset and want to do rigid registration.

What is the art of 3d multimodal rigid registration?

Example of the shape of the data:

The fixed image 512*512*197 and the moving images 512*512*497.

Hi.

I have a Lumencor Sola solid state light engine which uses a 5mm liquid light guide (LLG). I would like to use this for a spinning disk confocal setup which uses an FC port for optic fibres. I would like to stay away from lasers for now.

With some primitive calculations based on my objective lenses, I decided to purchase a 400um multimode (MM) optic fibre for UV-VIS (FC-SMA905 plugs). The Sola outputs IR as well, I have decided to either cut out that component with an IR cut filter or disconnect the LED module physically from the board. I do not think heat is good for the fibre.

The problem I am facing now is "squeezing" the output of the Sola into my 400um fibre. Realistically, an efficiency of 20% would be decent.

For the optical scheme, I basically plagiarised Thorlabs' solution for their stabilised light sources, which coincidentally also uses a 400um fibre bundle.

They appear to be using a 40mm best-form lens to collimate the output and an aspherical lens to focus it into the fibre.

I suppose the Lumencor Sola uses a similar method. I will have to open it and check, but I do recall a couple lenses being used, presumably to focus the light into the 5mm LLG. I do not wish to move those lenses and I also do a lot of widefield fluorescence imaging.

Therefore, I suppose I am attempting to collimate the output for a 5mm LLG and then focus it into my 400um MM fibre. I can design and 3D print a bracket for the Sola's output port which will enable a cage system for all the optics.

Another rather unusual method which I am unsure of would be focusing the output light with a microscope objective, straight from the Sola into the MM fibre.

Will my method(s) work? Is there a better method to achieve this with minimal alterations made to the Lumencor Sola?

Thank you for your help and any advice is appreciated!

We see a divergent beam of multimode KGW Raman laser with flat cavity mirrors, which is focused at the distance ~1.8 F, where F is lens focal length. It means that the beam has spherical component, which can be collimated by some lens at the output of Raman laser.

I think that it is well known effect, but can not find a paper, where this effect in Raman lasers is described. Can anybody give me a reference ?

Hi academics, I am looking for a journal that accepts papers on qualitative textual and multimodal discourse analysis of digital game dialogues. Discussion is related to social (and eco-)justice. Any recommendations? #linguistics #DigitalGames

We are trying to compare these two systems in the process of purchasing one of them. Our applications revolve around surface roughness quantification, mineral wettability evaluation, and surface force measurements on rock samples. I would appreciate your expert views.

Interpretable, credible and responsible multimodal artificial intelligence preface--DIKWP model (beyond ChatGPT)

Already 312 times read 2023-2-11 15:23 |System Classification: Paper Exchange

Subjects and objects in the entire digital world and cognitive world can be consistently mapped to the five components of the DIKWP model and their transformations: Data Graph, Information Graph, Knowledge Graph, Wisdom Map (Wisdom Graph), intention map (Purpose Graph).

Each DIKWP component corresponds to the semantic level of cognition, the concept and concept instance level of human language: {semantic level, {concept, instance}}

Model <DIKWP Graphs>

::=(DIKWP Graphs)*(Semantics, {Concept, Instance})

::={ DIKWP Graphs*Semantics, DIKWP Graphs*Concept, DIKWP Graphs*Instance }

::={ DIKWP Semanics Graphs, DIKWP Concept Graphs, DIKWP Instance Graphs }

Interactive scene <DIKWP Graphs>

::={DIKWP Content Graph includes: Data Content Graph, Information Content Graph, Knowledge Content Graph, Wisdom Content Graph, Purpose Content Graph;

DIKWP cognitive model (DIKWP Cognition Graph) includes: Data Cognition Graph, Information Cognition Graph, Knowledge Cognition Graph, Wisdom Cognition Graph , Purpose Cognition Graph.

}

Artificial intelligence is the capability part of DIKWP interaction.

AI::=(DIKWP Graphs)*(DIKWP Graphs)*

Narrow definition: Artificial intelligence is the development-oriented elimination of duplication in the DIKWP interaction, the integrated storage-computing-transmission iterative capability and the cross-DIKWP-oriented (Open World Assumption) OWA scope conversion capability.

We are conducting ChatGPT's artificial intelligence ability test and look forward to sharing it with you.

Examples of our related work:

Typed Resource-Oriented Typed Resource-Based Resource Management System (Authorized)

  Value Driven Storage and Computing Collaborative Optimization System for Typed Resources

   Application publication number: CN107734000A

For a list of all relevant invention patents, see:

List of Chinese national invention patents authorized by the DIKWP team for the first inventor Duan Yucong during the three years from 2019 to 2022 (69/241 in total)

Please visit: https://blog.sciencenet.cn/blog-3429562-1354842.html

Most polarization maintaining fibres available are single mode fibres. Does anyone know if any multimode polarization maintaining fibre products available? 

Edit: the paper was approved so if you want to see it just message me :)

I'm writing a paper on a multimodal active sham device for placebo interventions with electrostimulators. We believe it has a low manufacturing cost, but it's probably better to have some baseline for comparison. Have any of you ever requested a manufacturer to produce a sham replica of an electrostimulator to be used on blind trials? If so, how much did it cost? Was it an easy procedure?

Basically, I have read about mulsemedia, more precisely, the MPEG-V standard. However, also we have multimodal applications and a standard, W3C Multimodal Interaction Framework.

I'd like to know if these concept are antagonistic or they have similarities.

We have proposed an algorithm for multiobjective multimodal optimization problems and tested it on CEC 2019 benchmark dataset. We need to show the results on some real problem also. Kindly help.

And how can I download images from the whole brain ATLAS dataset provided by Harvard.

the website is http://www.med.harvard.edu/AANLIB/home.html, I can not find where to download.

Hi ! does anyone know if I can directly buy core only glass fibers ? Meaning no cladding or coating ?

I ideally am looking for a multimode glass core for biosensing purposes.

Hello there!

I have searched everywhere for a MRI dataset for amyotrophic lateral sclerosis, ideally a multimodal one (DTI especially would be appreciated).

Thank you in advance.

I formulated chitosan nanoparticles from a 0.5% w / v chitosan solution and 0.5% w / v TPP. after adding the TPP solution drop by drop to the Chitosan solution, I obtained a turbid suspension. I centrifuged at 3500 rpm for 30min and a pellet formed which I was able to resuspend with an ultrasound probe. After the size measurement at the DLS I have a multimodal distribution with most of the particles having a radius greater than 400nm.

The actual registration process is far from optimal as you can see from the attached picture.

Any idea on how to improve the registration process result?

What is the bandwidth specification of standard multimode OFC cables that are available for communications and networking? will they carry all UV-VIS-IR spectrum?

Dear All, within our new European project SYN+AIR related with the air transport we are executing an online survey which aims at identifing the mobility choices related to and from the airport. We are glad to invite you fill in the survey https://ec.europa.eu/eusurvey/runner/SYN_AIR_Traveller_Survey_2021 The questionnaire is available in 5 languages (English, Greek, Spanish, Italian, Serbian) and lasts approximately 10 minutes. All adults that travel or used to travel by plane (before the Covid-19 pandemics) can answer this survey. You may find information related to the project at http://syn-air.eu/

Please, feel free to share/disseminate this request. Thanks a lot for your attention and contribution. #SESAR #H2020 #SYN+AIR

We have datasets that have a Gaussian distribution.

,Data were obtained from different, irregular, and multimodal Gaussian distributions

How can we use the k-means clustering method for highly optimal clustering so that the most statistically similar data are in the same group?

We have seen a stability in the supply chains of goods, food in particular, during the current pandemic of Covid19 continue, mostly undisturbed.

It is very reassuring at a time of uncertainty and macro-risks falling onto societies.

How much do we owe to the optimised management and supervision of Container transport, and multimodal support to it with deep sea vessels, harbour feeder vessels, trains and trucks/lorries?

What is the granularity involved? Hub to hub, regional distribution, local delivery?

Do we think that the connectivity models with matrices, modelling the transport connections, the flows per category (passengers, freight, within freight: categories of goods), could benefit from a synthetic model agreggation of a single matrix of set federating what has been so far spread over several separate matrices of numbers?

What do you think?

Below references on container transport, and on matrices of sets

Hi everybody!

I am working on my diploma thesis regarding eye endoscope. I would like to know more about the speckles in the multimode fiber. I would like to reduce speckles in MM fiber using vibration. I do not know why the vibration reduce the speckles and what happens with modes that are in the optical fiber.

Thank you for your answer!

Regards

Barbora Spurná

Hi

I wanted to work on a comparison between the traditional class discourse interaction analysis and the new discourse version interaction system caused by the virus, and I also wanted to work on the part of the professors' opinions about the differences between these two discourse structures but I need some guidance to know poststructuralism or constructivism and in methodological frameworks, multimodal critical discourse analysis is the right one?. I have also doubted in the comparison that I should have one theory or draw results on two theories.

In advance thank you so much

Hi everybody!

I am working on my diploma thesis regarding eye endoscope. I would like to know more about the origin of speckles in the multimode fiber. I suppose that the speckles depend on fiber modes but I do not know why should high-order modes move with higher speed than low-order modes. And how does this fact influence the speckles.

Thank you for your answer!

Regards

Barbora Spurná

Accurate image captioning with the use of multimodal neural networks has been a hot topic in the field of Deep Learning. I have been working with several of these approaches and the algorithms seem to give very promising results.

But when it comes to using image captioning in real world applications, most of the time only a few are mentioned such as hearing aid for the blind and content generation.

I'm really interested to know if there are any other good applications (already existing or potential) where image captioning can be used either directly or as a support process. Would love to hear some ideas.

Thanks in advance.

What kind of matrices and why ?

I have to analyze construction project (it comprises images and text) from the viewpoint of multimodal analysis. Who has got some theoretical unformation about it or samples of multimodal analysis? Thanks a lot

I need to know if MM8 or MFP-3D origin gives more reliable data for the nanomechanical measurements like nanotube stiffness. Moreover, which one best performs in liquid and can give nanoparticle protein interaction forces. If anyone can tell me about vibration sensitivity of these that would be great also.

This dataset will be used in the context of a University Course.

Is it ok to say political caricatures instead of political cartoons in visual and multimodal metaphor?

Recently, I have been using the multimodal machine learning method to study the computer-aided diagnosis of cataract, but the data is not enough. Where can I find a multimodal data set? it is better to include image and structured data modalities.

I have a keystroke model which is one of the modes in my multimodal biometric system. The keystroke model gives me an EER of 0.09 using Manhattan Scaled Distance. But then I am normalizing this distance to fit in the range of [0, 1] using tanh normalization. And when I run a check on the normalized scores I am getting an EER of 0.997. Is there something I am doing wrong? The tanh normalization I am calculating based on the mean and std dev of matching scores for genuine users.

w/a = (0.65+(1.619./V.^(3/2)+2.879./V.^6)

Is this formula is also valid for step index multimoded waveguide for calculating the spot size of fundamental mode?

I am conducting research on Multimodal Discourse Analysis (MMDA) field. Which are the seminal works (books, papers, ...) in Multimodal Discourse Analysis (MMDA)?

I really appreciate knowing other researchers' point of view.

Thank you.

can we use score level fusion of Genuine and Imposter scores of a multi-Biomteric techniques to multimodal emotion recognition ?

I currently doing a research in fiber sensor using MMI, but i don't know how to determine the lenght of the multimode sensor. All reference that i read used the 4th self-image for determining the lenght of the multimode sensor, why the 4th self-image used ?

I'm doing some research on dimensionality reduction using swarm intelligent algorithms. As per the no free lunch rule, there is no algorithm that best suit all the problems. So, to be able to find the best subset I need to determine whether it's unimodal or multimodal? The data is of 300 features and 1000 instance. Is there any visualization methods that can help in this regard?

I made an experiment where I measure certain parameter a number of times. The result is over a 100 samples which distribution is not really normal. Due to properties of my specimens, results tend to gather around 3 or 4 modes. The distribution is multimodal. I would like to find the type A uncertainty of the measurement. When the distribution is normal, unimodal, the standard deviation is easily calculated. How to proceed when the distribution is multimodal? I found the stdv, same way as for unimodal, but I am not sure that this is correct way. Are there any dedicated standard deviation formulas for multimodal distribution? Even if I split my results into 3 or 4 separate unimodal sets, each with its own stdv, how to find the overall deviation?

I have known that the single-mode fiber (SMF) only allows the transmission of single-mode light, so we have to do mode-matching in SMF coupling. However, I am wondering what happens if we do not follow the strict mode coupling condition, we just couple all the energy into the SMF while do not follow the phase matching, how much loss will we get? Does the length of SMF have an influence on coupling efficiency in that situation?

Whether is it possible that we can still achieve less loss by a shorter SMF(with a range of few meters)?

In boundary condition ,it needs xmin xmax ymin ymax , should I put all of them pml or not?

And which size is enough for the FDE rectangular?

I am making a design where I have to splice PM single mode fiber to graded index multimode fiber, can it be possible with generic splicers or need some specialized one. what parameters should be considered to make an acceptable splice.

Thanks

In problems with many local optima (multimodal) and many variables to optimize (multidimensional) which PSO variants are those that provide:

It is grateful that in your response the bibliographic source where the PSO version is published is informed.

In (Wu, Y, et al 2017), the authors use 2 kind of objective functions (MI and DTV) with 4 optimization methods.

The attached table shows objective function result and RMSE. BTW, I can't understand how CLPSO with MI, DE with MI, ACO with MI and LMACO with MI have Mean and Best result for DTV? Since it runs with MI. If the authors says CLPSO, DE, ACO and LMACO without MI will be correct but they link the result with MI.

I hope my inquiry clear.

Dear Colleages,

I'm interested in visual and pictorial respresentations. Actaully, I have an idea that they have have a relationship with interextuality.

Please, there is any studies that focus on the intertextual analysis in the mutimodal (visual/pictorial) respresnetation, let me know.

Thanks

Hayder

I need some information on ATLAS.ti

Dear RG members,

Once I came across a software developed by Kay O Holloran to analyse moving pictures/videos. If you know such things, multimodal tools/models

I am simulation one fiber optic liquid level sensor where I am taking 1cm long multimode fiber. The cladding of the multimode fiber is removed by chemical etching process. For, measuring the liquid level, some portion of the fiber is immersed in the liquid and the remaining portion in in the air. Thus, the guided mode beam profile in the air-cladding section and that in the fluid-cladding section should be different. So there should be mode conversion loss.

Is there any theoretical formula to calculate such mode conversion loss.