Neural networks use electronic analogs of the neurons in our brains. But it doesn’t seem likely that just making enough electronic neurons would create a human-brain-like thinking machine. Consider that animal brains are sometimes larger than ours — a sperm whale’s brain weighs 17 pounds — yet we don’t think they are as smart as humans or even dogs who have a much smaller brain. MIT researchers have discovered differences between human brain cells and animal ones that might help clear up some of that mystery. You can see a video about the work they’ve done below.
Neurons have long finger-like structures known as dendrites. These act like comparators, taking input from other neurons and firing if the inputs exceed a threshold. Like any kind of conductor, the longer the dendrite, the weaker the signal. Naively, this seems bad for humans. To understand why, consider a rat. A rat’s cortex has six layers, just like ours. However, whereas the rat’s brain is tiny and 30% cortex, our brains are much larger and 75% cortex. So a dendrite reaching from layer 5 to layer 1 has to be much longer than the analogous neuron in the rat’s brain.
These longer dendrites do lead to more loss in human brains and the MIT study confirmed this by using human brain cells — healthy ones removed to get access to diseased brain cells during surgery. The researchers think that this greater loss, however, is actually a benefit to humans because it helps isolate neurons from other neurons leading to increased computing capability of a single neuron. One of the researchers called this “electrical compartmentalization.” Dig into the conclusions found in the research paper.
We couldn’t help but wonder if this research would offer new insights into neural network computing. We already use numeric weights to simulate dendrite threshold action, so presumably learning algorithms are making weaker links if that helps. However, maybe something to take away from this is that less interaction between neurons and groups of neurons may be more helpful than more interaction.
Most posts here are electrical or mechanical, with a few scattered hacks from other fields. Those who also keep up with advances in biomedical research may have noticed certain areas are starting to parallel the electronics we know. [Dr. Rajib Shubert] is in one such field, and picked up on the commonality as well. He thought it’d be interesting to bridge the two worlds by explaining his research using analogies familiar to the Hackaday audience. (Video also embedded below.)
He laid the foundation with a little background, establishing that we’ve been able to see individual static neurons for a while via microscope slides and such, and we’ve been able to see activity of the whole living brain via functional MRI. These methods gradually improved our understanding of neurons, and advances within the past few years have reached an intersection of those two points: [Dr. Shubert] and colleagues now have tools to peer inside a functional brain, teasing out how it works one neuron at a time.
[Dr. Shubert]’s talk makes analogies to electronics hardware, but we can also make a software analogy treating the brain as a highly optimized (and/or obfuscated) piece of code. Virus stamping a single cell under this analogy is like isolating a single function, seeing who calls it, and who it calls. This pairs well with optogenetics techniques, which can be seen as like modifying a function to see how it affects results in real time. It certainly puts a different meaning on the phrase “working with live code”!
It’s not too often that you see handkerchiefs around anymore. Today, they’re largely viewed as unsanitary and well… just plain gross. You’ll be quite disappointed to learn that they have absolutely nothing to do with this article other than a couple of similarities they share when compared to your neocortex. If you were to pull the neocortex from your brain and stretch it out on a table, you most likely wouldn’t be able to see that not only is it roughly the size of a large handkerchief; it also shares the same thickness.
The neocortex, or cortex for short, is Latin for “new rind”, or “new bark”, and represents the most recent evolutionary change to the mammalian brain. It envelopes the “old brain” and has several ridges and valleys (called sulci and gyri) that formed from evolution’s mostly successful attempt to stuff as much cortex as possible into our skulls. It has taken on the duties of processing sensory inputs and storing memories, and rightfully so. Draw a one millimeter square on your handkerchief cortex, and it would contain around 100,000 neurons. It has been estimated that the typical human cortex contains some 30 billion total neurons. If we make the conservative guess that each neuron has 1,000 synapses, that would put the total synaptic connections in your cortex at 30 trillion — a number so large that it is literally beyond our ability to comprehend. And apparently enough to store all the memories of a lifetime.
In the theater of your mind, imagine a stretched-out handkerchief lying in front of you. It is… you. It contains everything about you. Every memory that you have is in there. Your best friend’s voice, the smell of your favorite food, the song you heard on the radio this morning, that feeling you get when your kids tell you they love you is all in there. Your cortex, that little insignificant looking handkerchief in front of you, is reading this article at this very moment.
What an amazing machine; a machine that is made possible with a special type of cell – a cell we call a neuron. In this article, we’re going to explore how a neuron works from an electrical vantage point. That is, how electrical signals move from neuron to neuron and create who we are.
A Basic Neuron
Despite the amazing feats a human brain performs, the neuron is comparatively simple when observed by itself. Neurons are living cells, however, and have many of the same complexities as other cells – such as a nucleus, mitochondria, ribosomes, and so on. Each one of these cellular parts could be the subject of an entire book. Its simplicity arises from the basic job it does – which is outputting a voltage when the sum of its inputs reaches a certain threshold, which is roughly 55 mV.
Using the image above, let’s examine the three major components of a neuron.
The soma is the cell body and contains the nucleus and other components of a typical cell. There are different types of neurons whose differing characteristics come from the soma. Its size can range from 4 to over 100 micrometers.
Dendrites protrude from the soma and act as the inputs of the neuron. A typical neuron will have thousands of dendrites, with each connecting to an axon of another neuron. The connection is called a synapse but is not a physical one. There is a gap between the ends of the dendrite and axon called a synaptic cleft. Information is relayed through the gap via neural transmitters, which are chemicals such as dopamine and serotonin.
Each neuron has only a single axon that extends from the soma, and acts similar to an electrical wire. Each axon will terminate with terminal fibers, forming synapses with as many as 1,000 other neurons. Axons vary in length and can reach a few meters long. The longest axons in the human body run from the bottom of the foot to the spinal cord.
The basic electrical operation of a neuron is to output a voltage spike from its axon when the sum of its input voltages (via its dendrites) crosses a specific threshold. And since axons are connected to dendrites of other neurons, you end up with this vastly complicated neural network.
Since we’re all a bunch of electronic types here, you might be thinking of these ‘voltage spikes’ as a difference of potential. But that’s not how it works. Not in the brain anyway. Let’s take a closer look at how electricity flows from neuron to neuron.
Action Potentials – The Communication Protocol of the Brain
The axon is covered in a myelin sheet which acts as an insulator. There are small breaks in the sheet along the length of the axon which are named after its discoverer, called Nodes of Ranvier. It’s important to note that these nodes are ion channels. In the spaces just outside and inside of the axon membrane exists a concentration of potassium and sodium ions. The ion channels will open and close, creating a local difference in the concentration of sodium and potassium ions.
We all should know that an ion is an atom with a charge. In a resting state, the sodium/potassium ion concentration creates a negative 70 mV difference of potential between the outside and inside of the axon membrane, with there being a higher concentration of sodium ions outside and a higher concentration of potassium ions inside. The soma will create an action potential when -55 mV is reached. When this happens, a sodium ion channel will open. This lets positive sodium ions from outside the axon membrane to leak inside, changing the sodium/potassium ion concentration inside the axon, which in turn changes the difference of potential from -55 mV to around +40 mV. This process in known as depolarization.
One by one, sodium ion channels open along the entire length of the axon. Each one opens only for a short time, and immediately afterward, potassium ion channels open, allowing positive potassium ions to move from inside the axon membrane to the outside. This changes the concentration of sodium/potassium ions and brings the difference of potential back to its resting place of -70 mV in a process known as repolarization. Fro start to finish, the process takes about five milliseconds to complete. The process causes a 110 mV voltage spike to ride down the length of the entire axon, and is called an action potential. This voltage spike will end up in the soma of another neuron. If that particular neuron gets enough of these spikes, it too will create an action potential. This is the basic process of how electrical patterns propagate throughout the cortex.
The mammalian brain, specifically the cortex, is an incredible machine and capable of far more than even our most powerful computers. Understanding how it works will give us a better insight into building intelligent machines. And now that you know the basic electrical properties of a neuron, you’re in a better position to understand artificial neural networks.
Artificial Intelligence is playing an ever increasing role in the lives of civilized nations, though most citizens probably don’t realize it. It’s now commonplace to speak with a computer when calling a business. Facebook is becoming scary accurate at recognizing faces in uploaded photos. Physical interaction with smart phones is becoming a thing of the past… with Apple’s Siri and Google Speech, it’s slowly but surely becoming easier to simply talk to your phone and tell it what to do than typing or touching an icon. Try this if you haven’t before — if you have an Android phone, say “OK Google”, followed by “Lumos”. It’s magic!
Advertisements for products we’re interested in pop up on our social media accounts as if something is reading our minds. Truth is, something is reading our minds… though it’s hard to pin down exactly what that something is. An advertisement might pop up for something that we want, even though we never realized we wanted it until we see it. This is not coincidental, but stems from an AI algorithm.
At the heart of many of these AI applications lies a process known as Deep Learning. There has been a lot of talk about Deep Learning lately, not only here on Hackaday, but all over the interwebs. And like most things related to AI, it can be a bit complicated and difficult to understand without a strong background in computer science.
If you’re familiar with my quantum theory articles, you’ll know that I like to take complicated subjects, strip away the complication the best I can and explain it in a way that anyone can understand. It is the goal of this article to apply a similar approach to this idea of Deep Learning. If neural networks make you cross-eyed and machine learning gives you nightmares, read on. You’ll see that “Deep Learning” sounds like a daunting subject, but is really just a $20 term used to describe something whose underpinnings are relatively simple.
When you want a person to do something, you train them. When you want a computer to do something, you program it. However, there are ways to make computers learn, at least in some situations. One technique that makes this possible is the perceptron learning algorithm. A perceptron is a computer simulation of a nerve, and there are various ways to change the perceptron’s behavior based on either example data or a method to determine how good (or bad) some outcome is.
What’s a Perceptron?
I’m no biologist, but apparently a neuron has a bunch of inputs and if the level of those inputs gets to a certain level, the neuron “fires” which means it stimulates the input of another neuron further down the line. Not all inputs are created equally: in the mathematical model of them, they have different weighting. Input A might be on a hair trigger, while it might take inputs B and C on together to wake up the neuron in question. Continue reading “Machine Learning: Foundations”→
If one hack that controls amputated cockroach legs this week wasn’t enough for you, we’ve got another.
Earlier this week we saw two neuroscientists at Backyard Brains put on a show at a TED talk by connecting an amputated cockroach leg (don’t worry, they grow back) to a $100 electronic device called the SpikerBox. The SpikerBox allows students to explore the world of axons and action potentials by listening in on the electronic signals generated by the hair on the legs of a cockroach. For the finale for their TED talk, the SpikerBox guys attached an MP3 player to the cockroach leg, causing the now dead appendage to dance a little jig.
This new build – the Salt Shaker from Thinker Thing again allows students to amputate cockroach legs, pin them down with electrodes, and cause muscle contractions with the sound of science. Thinker Thing took this one step further than the neuroscientists at Backyard Brains; now you can control a cockroach leg with your mind.
The folks at Thinker Thing are using an off the shelf EEG system from Emotiv to capture the alpha, beta, and delta brainwaves of their new human test subjects. By interpreting these brain signals, they can convert these small variations in cerebral electrical activity to sound files. From there, it’s simply a matter of plugging in the Salt Shaker and moving a cockroach leg with your mind.
In the video after the break you can check out the folks at Thinker Thing playing around with their Salt Shaker and controlling a cockroach leg with a team member’s mind.
Neuroscientists [Tim Marzullo] and [Greg Gage] wanted a way to get kids interested in neuroscience. What they came up with isn’t terribly far from something found in Frankenstein’s lab; by amputating a cockroach’s leg and attaching electrodes, they’re able to listen to the sound of neurons firing. For an even cooler demonstration, they’re able to apply a little bit of current to the leg and make the leg dance to the beat of the Beastie Boys.
The guys published an article in PLOS One and gave a TED talk demonstrating their SpikerBox, as they call their invention, to the masses. The basic idea is to amplify the electricity generated by cockroach neurons firing. By listening in on the neurons with an iPad app, [Tim] and [Greg] can set the threshold of the recording to detect the action potential of an individual neuron, and listen in on exactly what happens when a single neuron fires.
It seems like a great tool to explain the very basics of what a nervous system – and a brain, both cockroach and human – actually is. In the video after the break, you can see [Greg] playing around with individual cockroach neurons. After that, [Greg] plays the Beastie’s High Plains Drifter into the leg making the muscles contract. Truly, The Sounds of Science.