Home | Community | Message Board

NorthSpore.com BOOMR Bag!
This site includes paid links. Please support our sponsors.


Welcome to the Shroomery Message Board! You are experiencing a small sample of what the site has to offer. Please login or register to post messages and view our exclusive members-only content. You'll gain access to additional forums, file attachments, board customizations, encrypted private messages, and much more!

Shop: Left Coast Kratom Buy Kratom Extract   Kraken Kratom Red Vein Kratom   North Spore Bulk Substrate   Mushroom-Hut Substrate Bags   Original Sensible Seeds Bulk Cannabis Seeds   PhytoExtractum Maeng Da Thai Kratom Leaf Powder

Jump to first unread post Pages: 1
OfflinePTreeDish
Male


Registered: 04/22/18
Posts: 353
Last seen: 4 months, 3 days
Machine Learning Applications in Mycology
    #26506874 - 02/27/20 05:09 PM (3 years, 11 months ago)

I know how to code and do some basic machine learning to make predictions on large volumes of labelled data. I'm curious if anyone here knows of any existing machine learning applications being used in mycology? Presumably, ML would be useful in identifying new sp. or evaluating the genome.

I'd also love to hear any ideas about what any machine learning applications you think might be useful to the community. Maybe I'll take this on as a side project and built it.


Extras: Filter Print Post Top
OfflineeLeSDenes
Mycelium Expander
Male User Gallery


Folding@home Statistics
Registered: 05/25/11
Posts: 955
Last seen: 2 years, 7 days
Re: Machine Learning Applications in Mycology [Re: PTreeDish] * 1
    #26506948 - 02/27/20 05:49 PM (3 years, 11 months ago)

Machine learning is used a lot in biology for example in recognising cancerous cells.
you could use binary classifiers to tell whether a substrate is contaminated from images. If you get enough labelled data is fairly easy to do.
The hard thing is not the prediction, it is to get good quality training data.


Extras: Filter Print Post Top
OfflinePTreeDish
Male


Registered: 04/22/18
Posts: 353
Last seen: 4 months, 3 days
Re: Machine Learning Applications in Mycology [Re: eLeSDenes]
    #26506978 - 02/27/20 06:07 PM (3 years, 11 months ago)

That is correct re: good input data. Is there sufficient value in automatically detecting contamination from a plate though? I’m sure it would be useful to newcomers to have a website where they can upload an image of their plate and it shows them bounding box on their image with a % confidence where possible contam is detected but I would think new users would want to learn to identify these themselves.

That said, a few of us could probably scrape and label known contaminated plates from shroomery post so that I could build a reasonably decent tool. I just wonder if that particular use case is worth the effort?


Edited by PTreeDish (02/27/20 06:08 PM)


Extras: Filter Print Post Top
OfflineeLeSDenes
Mycelium Expander
Male User Gallery


Folding@home Statistics
Registered: 05/25/11
Posts: 955
Last seen: 2 years, 7 days
Re: Machine Learning Applications in Mycology [Re: PTreeDish]
    #26507001 - 02/27/20 06:18 PM (3 years, 11 months ago)

That was just a bold idea :laugh: not sure if it is useful probably it is just fun to do. A more useful thing would be to identify if a substrate will have contamination or not. There are early signs which might not always be obvious


Extras: Filter Print Post Top
OfflinePTreeDish
Male


Registered: 04/22/18
Posts: 353
Last seen: 4 months, 3 days
Re: Machine Learning Applications in Mycology [Re: eLeSDenes]
    #26507133 - 02/27/20 07:43 PM (3 years, 11 months ago)

I think its an interesting idea! Thank you for even contemplating my question. šŸ™šŸ»


Extras: Filter Print Post Top
Offlinegone-pear-shaped
Stranger than fiction

Registered: 10/30/17
Posts: 822
Last seen: 6 months, 4 days
Re: Machine Learning Applications in Mycology [Re: PTreeDish]
    #26507361 - 02/27/20 10:28 PM (3 years, 11 months ago)

Regarding the value of automated contam classification, I'm afraid the value would lie in quickly classifying a large amount of inputs which could later be given attention by humans to confirm. I could see that being useful for enormous mushroom farms--are there any enormous mushroom farms?--but for the type of person this site is targeted at, would it give more value to the individual and the community than if the user posted in the 'contamination' forum?

What about automated classification of mushroom variety?

And I don't have specific ML ideas regarding the following, but the current (planning) problems I'm pondering are how to match spores/agar recipe/spawn/substrate with the locally available materials, what I already have on hand, my climate, what I expect the weather to do in the next couple months. But I feel all that is more suited to an expert system rather than machine learning.


Extras: Filter Print Post Top
Offlinedark-goblin
Stranger


Registered: 01/16/20
Posts: 13
Last seen: 1 year, 10 months
Re: Machine Learning Applications in Mycology [Re: gone-pear-shaped]
    #26512609 - 03/02/20 10:50 AM (3 years, 11 months ago)

seems like the problem is the training data, as user above stated. i can think that image training is only useful if you have a definitive genetic sequence ID to pair with each image, and as far as im aware + from what i understand about neural network training, you'd need a lot of images for that kind of thing.

it's an interesting application, and a bold proposal, which given enough time (years of effort) and dedication there would surely eventually be some fruitful outcome. but yes the first problem is deciding on some useful data sets + then acquiring said data.

this thread might be useful for ideation in that regard, would also be interested to hear what ideas other people have


Extras: Filter Print Post Top
OfflinePTreeDish
Male


Registered: 04/22/18
Posts: 353
Last seen: 4 months, 3 days
Re: Machine Learning Applications in Mycology [Re: dark-goblin]
    #26512654 - 03/02/20 11:07 AM (3 years, 11 months ago)

For sure, training data is a part of creating a solution, but I wouldn't even worry about that until a good use-case for ML is evaluated first. Once we find a good application of the tech, then we can work on how to solve the problem.

Quote:

I could see that being useful for enormous mushroom farms--are there any enormous mushroom farms?--but for the type of person this site is targeted at, would it give more value to the individual and the community than if the user posted in the 'contamination' forum?





Even commercial farms probably wouldn't employ ML unless it offered enough return to justify the expense of changing their existing processes. And for contam ID, their finely tuned eyes are probably faster than any other tech. Of course, if there was a massive assembly line and a camera over the top of the conveyor was able to take pics and trigger an action (like blowing the defective grow off the main belt), that might be useful - but I can't imagine anyone is growing at this kind of scale.


Quote:

i can think that image training is only useful if you have a definitive genetic sequence ID to pair with each image, and as far as im aware + from what i understand about neural network training, you'd need a lot of images for that kind of thing.




Just a few thoughts and questions on your comments here. Why would a DNA sequence be needed to pair with each image?

Also, just FYI, a neural network is only one form of applying ML, but not necessarily always the best approach to all problems. Also, the need for lots of images depends on how large the standard deviation is for the images being examined.

If it were random images being submitted by users on the forum, yes - that would require a lot of training data since the permutations vary wildly. But if the predictions were based on images made in a controlled setting, then I've seen some applications work quite well with as little as 20 images.

Quote:

it's an interesting application, and a bold proposal, which given enough time (years of effort) and dedication there would surely eventually be some fruitful outcome. but yes the first problem is deciding on some useful data sets + then acquiring said data.




I'm glad that you see the potential! I would just counter what you said with trying to focus on a useful application first and then we can work together to figure out how to get the training data (which I know may have huge implications on the feasibility), but in my experience, when it comes to these kinds of brainstorming exercises, we can often limit our imagination by prematurely limiting ourselves out of fear that we may have an idea that might not be possible.

I'm actually curious to hear, if we could imagine that anything is possible for a moment, what application of ML would be a real game changer for the field of mycology.

So glad to see so many fellow coders here!


Extras: Filter Print Post Top
Jump to top Pages: 1

Shop: Left Coast Kratom Buy Kratom Extract   Kraken Kratom Red Vein Kratom   North Spore Bulk Substrate   Mushroom-Hut Substrate Bags   Original Sensible Seeds Bulk Cannabis Seeds   PhytoExtractum Maeng Da Thai Kratom Leaf Powder


Similar ThreadsPosterViewsRepliesLast post
* [STICKY] Advanced Mycology Forum Guidelines Raadt 37,873 0 04/07/03 08:05 PM
by Raadt
* Help me fuffill my mycological goals! Fuzzy 2,965 7 05/18/01 07:28 AM
by tom
* bulk straw..... washing machine idea!
( 1 2 all )
Curious_George 4,590 22 03/11/14 10:10 PM
by senorcafe
* Applications of mycelium piss?
( 1 2 all )
TheShroomHermit 3,204 28 09/08/05 03:59 PM
by shirley knott
* mycology derx 2,152 2 02/27/05 08:40 PM
by Joops
* Korean Journal of Mycology aaron 2,640 1 03/07/02 09:55 AM
by mycofile
* FOG machine
( 1 2 all )
SixTango 4,457 28 03/15/08 04:48 PM
by Familiar Gnome
* BIG DADDY WET PASTURIZING MACHINE: agar 2,624 13 12/30/04 10:32 PM
by jeff01

Extra information
You cannot start new topics / You cannot reply to topics
HTML is disabled / BBCode is enabled
Moderator: RogerRabbit, Pastywhyte, bodhisatta
605 topic views. 1 members, 7 guests and 1 web crawlers are browsing this forum.
[ Show Images Only | Sort by Score | Print Topic ]
Search this thread:

Copyright 1997-2024 Mind Media. Some rights reserved.

Generated in 0.025 seconds spending 0.004 seconds on 12 queries.