AI Let’s reveal in terms of biomedical
Separate the words “Artificial” and “Intelligence”. Both words are so simple to understand. But here intelligence is in brief, which regards to brain’s functionality, behavior, memorization, signal processing, information processing with neurons and learning power. Intelligence is the ability to acquire and apply knowledge and skills.
As an IT engineer, if someone asks me about how and where to use AI, I can say that we can use machine learning frameworks and platforms like TensorFlow, PyTorch, CAFFE, Keras, Chainer and so on for making software and applications i.e. Face recognition, Image classification, stock market prediction, chess game, etc. and will say that It’s generally subject of data scientists.
But basically, the core principles of AI(ML) belongs to Biomedical. I mean the basic principle is related to the human brain and its behavior. Based on that principle data scientists develop such algorithms using vast math subjects like vector, tensors, matrices, etc. for making machine learning models and train them.
We will see what biomedical students learn about AI? How is it related to Information and technology. Let’s see.
Human Senses
Human has five senses touch, smell, vision, voice recognition, and taste. Which all are generating information for the brain according to their functionalities.
When that information is generated it will be memorized once. Then again if the brain will get this information with senses it will automatically be memorized.
Let’s take an example of a vision. If 5–6-month baby saw an apple for the first time in life and we taught him it’s an apple. Now that apple’s image information is stored in his memory. So, the second time if we show apple to him, his brain will start the process for memorizing and get the output that it is apple. The time for generating that information is based on how much accurately that information(Data) is already memorized by the brain. So, now if he saw apple third time, the information generating will be more accurate and faster than the second time and so on. So, we can say that the brain is learning. That same principle can be used in deep learning and machine learning.
Neurons
Neurons are specialized cells that handle the generation, calculation, and transmission of information to the brain. First input came from any of the senses then neurons make that path to the brain. This type of neurons is called Sensory Neurons. They respond to stimuli as input such as touch, sound, or light that affect the cells of the sensory organs, and they send signals to the brain. The Biological Neuron looks like this.
Where dendrites working as inputs. We can see here there is more than one input. Axon working as an output. Axon terminals will feed output to the next neuron. It’s also called a synapse. There can be also so many outputs. The cell body and nucleus will calculate and generate information in terms of electrical signals. When neurons are passing information one to one they are passing electrical signals generated by cell fluid. That’s enough for the basic understanding of biological neurons.
Ahh!! But that’s biology. Where is this concerned with AI or machine learning?
Then see this mathematical model of a biological neuron.
We can correlate easily that model with the biological neuron. There are so many neurons, connected with axons and dendrites and make a network for information passing, which is called a neural network.
We are not talking about calculation or equation here. In the image, it’s just indicated for showing that, there is some kind of process for information generation with natural neuron cell, which can be artificially understood and derived by some equation in the real world.
A neural network is a very familiar term for machine learning. If we compare natural and artificial neural networks it looks like this.
We could understand now, that how any information from senses reaches to the brain.
But still, we have other questions, why are we using this concept for machine learning? How machine will learn with this kind of structure and calculation?
That question arises because we are missing one most important natural phenomenon of neurons. This is called Neuronal memory allocation. It’s a part of sensory memory.
Sensory Memory
Sensory memory is the shortest-term element of memory. It is the ability to retain impressions of sensory information after the original stimuli have ended. It acts as a kind of buffer for stimuli received through the five senses of sight, hearing, smell, taste, and touch, which are retained accurately, but very briefly.
For example, the ability to look at something and remember what it looked like with just a second of observation is an example of sensory memory.
In summary we can say that memory is stored some ware in side neuron and this process is called Neuronal memory allocation.
Neuronal allocation is a phenomenon that says how specific neurons in a network storing a specific memory and not others that receive similar input, which is committed to.
We could understand that same from our apple’s recognition example.
If we correlate this to machine learning, it has the same functionality. It is clear that first, we have to train some kind of a bunch of neural networks, we can say it, ML model.
Then we can use that model file in some ware in application or software. So, that ML model is specifically memorizing some predefined required set of data, and based on its processing output.
So, after understanding neuronal functionalities we could understand how AI(ML) is related to concepts of biomedical.
Thanks for reading this article. I like to hear feedback from you. Feel free to ask any questions :).
