Cellular neural networks

[IRobot_129]

I am writing my thesis on robot path planning, I found Cellular Neural Networks fascinating, it is simple yet powerful like cellular automata. Is CNN still used in real-world applications or dead in machine learning? I would love insights from someone more experienced

 

[Manny J.]

I am writing my thesis on robot path planning, I found Cellular Neural Networks fascinating, it is simple yet powerful like cellular automata. Is CNN still used in real-world applications or dead in machine learning? I would love insights from someone more experienced

Yes, CNN is still of course used in real-world applications and even increasingly so... now that we're approaching total modernization or something. It's underrated gold if you ask me.

 

[EnduranceEnthusiast]

I am writing my thesis on robot path planning, I found Cellular Neural Networks fascinating, it is simple yet powerful like cellular automata. Is CNN still used in real-world applications or dead in machine learning? I would love insights from someone more experienced

Cellular Neural Networks are not "dead," but they occupy a specialized niche, particularly in hardware and specific real-time signal processing.
While Deep Learning (especially DRL with digital CNNs) dominates current robot path planning research, your fascination with CNNs offers a unique angle. By focusing on their inherent parallelism, local computation, and potential for efficient hardware implementation, you can make a strong case for their continued relevance and potentially explore novel hybrid solutions for robot path planning for your thesis, just a thought tho

 

[Capturingcaptcha]

To add some additional info that both of you outlined, I read a while ago that there are research companies in eu that are still developing cnn based chips for certain applications. It's quite impressive how powerful cnn actually is, despite being technically replaced by convolutional neural networks

 

[abcdefu]

CNNs in the cellular sense aren't exactly mainstream anymore, but they're not dead either. I think a lot of niche hardware-focused apps still swear by them. If it fits your case, then why not?

 

[IRobot_129]

Yes, CNN is still of course used in real-world applications and even increasingly so... now that we're approaching total modernization or something. It's underrated gold if you ask me.

Thanks! I was actually referring to Cellular Neural Networks, not Convolutional ones. Do you think CNNs (the cellular type) still have relevance in modern robotics or embedded systems?

 

[IRobot_129]

Cellular Neural Networks are not "dead," but they occupy a specialized niche, particularly in hardware and specific real-time signal processing.
While Deep Learning (especially DRL with digital CNNs) dominates current robot path planning research, your fascination with CNNs offers a unique angle. By focusing on their inherent parallelism, local computation, and potential for efficient hardware implementation, you can make a strong case for their continued relevance and potentially explore novel hybrid solutions for robot path planning for your thesis, just a thought tho

I hadn't thought about hybrid solutions before, what kinds of hardware platforms would best suit CNNs for real-time path planning in robotics, in your opinion?

 

[IRobot_129]

To add some additional info that both of you outlined, I read a while ago that there are research companies in eu that are still developing cnn based chips for certain applications. It's quite impressive how powerful cnn actually is, despite being technically replaced by convolutional neural networks

Thanks!