Smart Nose


#1

Related to: [Smart Nose]
Hi All,
Not sure weather I am writing on correct place or not.
I am always wondering is it possible we can save and calculate smell using our computing machines. Can we train machines to identifying an odour and create one for me.
For this I tried reading about how our nose works ?
I got to know it goes from our nose and then olfactory gland take care of it which help in distinguishing between various type of odours… this is an short introduction of how a nose works .
But to make it in artificial way I tried to research on how odours are make, and I went into organic chemistry and tried to learn more how do it make smells , there reactions etc etc

I am not good at chemistry but found it feisible how smells are changed from bad odour to pleasant smell. It all happens due to a increment in hydrogen and carbon chain.

Uploading…

I am very interested in more research but it is not sure where to go, how will it come in the picture of computing.

Hope some of you may be knowledgeable over organic reactions and chemistry to compute :grinning::yum:
Any ideas??

Thanks,
Rahul


#2

Hi Rahul,

You may want to check out Professor Barani Raman’s work on machine olfaction:

Professor Tim Holy may also have work related to your interests:

http://dbbs.wustl.edu/faculty/Pages/faculty_bio.aspx?SID=5568

Let me know if this is relevant to your ideas and I might be able to help,

Charlie


#3

Hi Charlie,
First of all , Thanks for looking in to this :slight_smile:
You comment gave me new way to learn , I was searching for the same kind of stuff, Similar to it. but my first intension is to differ between good and bad smell.
By your link Prof. Barani Raman’s work , i got a way to understand this process. I got a handbook too Machine olfaction
I will go through it , and will explore some more intresting studies :slight_smile:

On other side :
I got some clues with VSEPR Theory,
I could say , It is the one which make the chemical bonds with some flavour smell. I am still studying how to proof it using Quantum mechanics formulae.

You can help with your feedbacks and suggestions , it will be my pleasure :slight_smile:


#4

Hi Rahul,

I am not an expert in all these fields, so I do not feel too comfortable adding much more. My only suggestion would be to be cautious of how much material you are working on.

One approach is through neural coding and machine learning and statistics. I suspect that may be the most straightforward approach for what you are working on?

I recommend against a more chemistry-based approach for now if you are not confident in your ability in this area. VSEPR barely scratches the surface of what you might need.

There are competing theories of olfaction. The newer one uses arguments involving quantum mechanics, so you are on track, but there’s a lot of work that’s still ongoing.

And of course there is some recent work in computational chemistry and machine learning that you may like (https://arxiv.org/abs/1701.04503) but again, it’s a different direction …

I feel that is a lot of high-level material across many disciplines, so depending on the application you might want to find professors in the area you are most interested first.

Charlie


#5

Hi Charlie,

My First intention is to make my application or a machine to understand what is smelling bad and what is good.

You are absolutly right , the best way is to go through neural coding , regressions charts and then make machine learn what i want. This is what i will also do. But before doing it i need to understand olfactions, and to understand it i went till VSEPR theory.

Well I am a Software Engineer in Embedded domain and this is my part time research and i am very interested to make it done. I will read about computational chemistry , But still first i have to understand how will i be calculating it using quantum mechanics.

Yes , i need some expertise people to whom i can ask question, but looks like in my part time it is very hard to find some one on this. :wink: So i found this forums to get more precious ideas/thoughts/suggestions/theories like you shared.

Is there any open way to do this , i mean like a online forum (like Open AI) or some web site who provide user to research freely.? I am a big fan of Opensource hence i want this study to be Open to all , any one can put ideas, contribute and use as they want.

Thanks,
Rahul


#6

Hi Rahul,

If your first intention is to make an application, I agree with the neural coding approach. The problem is getting the data, so I still recommend finding a professor.

I think I understand why you might look into chemistry-based approaches. Fortunately there are online forums for this. Unfortunately you might have to go to many places.

Again I am not an expert, but if I were to start, I might look into quantum chemistry and computational chemistry software first, then find people working on specific projects.

Here is a starting point: https://en.wikipedia.org/wiki/List_of_quantum_chemistry_and_solid-state_physics_software

And here are some online forums that may be helpful:
PSI4 (quantum chemistry package): http://forum.psicode.org/

iOpenShell: http://iopenshell.usc.edu/forum/, discussions may help: http://iopenshell.usc.edu/forum/topic.php?id=11&usebb_sid=967acab6ffda8eb4e806ebfc2d2b71ca

NWChem (quantum chemistry package): http://www.nwchem-sw.org/index.php/Special:AWCforum

Though again I might still recommend searching out specific groups that may have professors (more likely students) that can help you, maybe: https://dasher.wustl.edu/

Charlie


#7

Hi Rahul,

Happened across more recent work in computational chemistry and machine learning by Google that you may find interesting or useful here (the two papers mentioned in this link):

Hope this is useful,

Charlie


#8

Hello Charlie,
Sorry for late reply , I was engaged with some other work this week.
Your last reply was really very helpful , Thanks for it :wink:
But as you suggested https://dasher.wustl.edu/ I email to professor jay regarding the study, may be if he find that intresting he will respond to it.

And i am continuing with the handbook of Machine Olfaction.
It have many of the thing which i was curious to know. I will try to complete it fast and will make some Regressions.

But how will i get a machine which can inhale ?? Like Our nose. Any suggestions for it ?

Looking for more People , if can join in this research …! :slight_smile:

Thanks,
Rahul


#9

Hi Rahul,

  1. Sorry for my late reply as well, I have not been checking this thread often.

  2. I wouldn’t count on professors replying, they are often very busy. Don’t be afraid to contact the students as well.

  3. For a machine that can inhale, I mentioned a professor earlier you might have missed - Professor Tim Holy - he had a student who studied nasal aerodynamics and how that contributes to smell - very relevant to your interests I hope. I believe the student even 3D printed a nose with a specific shape to test her theories. If you are interested you may try to contact the professor or I might be able to ask for more information.

Hope your research goes well,

Charlie