More on this book
Kindle Notes & Highlights
bifurcations mark a quick change between two separate areas in parameter space. In
Reduce the network’s input and it oscillates more slowly, around 20 times a second.
Groups of neurons in the visual cortex, for example, can oscillate at a swift 60 times a second.
High-frequency oscillations are referred to by neuroscientists as ‘gamma’ waves.
The reason for this is that Hans Berger, the inventor of the original EEG machine, called the big slow waves he could see by eye on his shoddy equipment ‘alpha’ waves and everything else ‘beta’;
Gamma waves, while fast, are usually small – or ‘low amplitude’ in technical terms. Their presence, detectable by a modern EEG or an electrode in the brain, is associated with an alert and attentive mind.
It’s as though the voice neurons, by being the first to fire, are pushing themselves
through a door and then slamming it shut on the background neurons. And
how oscillations go wrong in diseases like schizophrenia, bipolar disorder and autism.
There are other equally respected neuroscientists that, colourfully, view gamma rhythms as the exhaust fumes of computation: They show up when the engine is running but they’re absolutely not important.’
widespread assumption that women in math were like dancing bears – perhaps they could do it, but not very well, and the attempt was an amusing spectacle.’
Oliver Selfridge described the process of template matching at a 1958 conference.
‘We are not going to apologise for a frequent use of anthropomorphic or biomorphic terminology. They seem to be useful words to describe our notions.’
hierarchical. That is,
Pandemonium breaks the problem of vision into two stages: first look for the simple things, then for the more complex.
Inside the retina, individual light-detecting photoreceptors send their information to another population of cells called ganglion cells.
each ganglion cell gets inputs from many photoreceptors.
crucially, all these inputs come from a certain limite...
This highlight has been truncated due to consecutive passage length restrictions.
This makes a ganglion cell responsive only to light that hits the retina in that specific location – and each cel...
This highlight has been truncated due to consecutive passage length restrictions.
1959 paper ‘What the frog’s eye tells the frog’s brain’
David Hubel6 and Torsten Wiesel focused on was the primary visual cortex. This
strange: occasionally the neuron would respond – not to the slides – but to the changing of them.
‘In a certain early phase of science a degree of sloppiness can be a huge advantage.’
dug deep into how this responsiveness to moving lines worked.
neurons in the primary visual cortex each have a preferred orientation in addition to a preferred location.
Flies have bug detectors and cats (and other mammals) have line detectors. However,
cells they get their inputs from – those in the thalamus – respond to dots, not lines.
neurons in the primary visual cortex listen for the activity of neurons in the thalamus that make up their preferred line.
These neurons would respond if a line appeared anywhere in a region that was about four times larger than that of the other neurons they recorded.
how points of light become perception.
‘Our idea originally was to emphasise the tendency toward increased complexity as one moves centrally along the visual path, and the possibility of accounting for a cell’s behaviour in terms of its inputs.’
simple cell is a line with a specific orientation. In engineering terms, this grid of numbers is known as a filter
the activity of one simple cell was calculated as the
sum of the thalamus activity at one location multiplied by the filter.
Sliding the filter across the whole image created a set of simple cells all with the same preferred orientation but different preferred locations. This is a p...
This highlight has been truncated due to consecutive passage length restrictions.
This second layer of simple cells would be on the lookout not for simple features in the image, but rather for simple ‘features’ in the activity of the complex cells from which they get their input. They’d
The spikes of a neuron are thus like an army of ants – each as identical as possible, their power coming mainly from their numbers.
MacKay and McCulloch calculated an upper bound of 2,900 bits per second.
Shannon estimated the redundancy of written English to be about 50 per cent.
while keeping the weight applied to the muscle constant, the firing rate of the nerve would decrease by about half over 10 seconds. Adrian called this phenomenon ‘adaptation’ and defined it as ‘a decline in excitability caused by the stimulus’.
neurons in the motor cortex on one side of the brain send outputs to neurons in the spinal cord on the other side, and these spinal cord neurons go directly to specific muscle fibres.
neuron fires, it releases the neurotransmitter acetylcholine into this junction. Muscle fibres respond to acetylcholine by contracting and movement occurs. Through this path, neurons in the cortex can directly control muscles.
brainstem, basal ganglia and cerebellum.
there are many channels through which the cortex can communicate with the muscles and many possible messages that could be sent.
cortex might do something useful still assumed that it worked as an undifferentiated mass – a web of tissue without functional specialisation.
Different portions of motor cortex work like keys on a piano, each producing its own single note.
short melodies
This argument – notes versus melodies, twitches versus movements – would be the first of many debates over the soul of the motor cortex.
Thus, the fashion became to try to stimulate as small an area of the cortex as possible to see just how localised the function could get. This trend aligned very well with Fritsch and Hitzig’s approach of using short small pulses to cause individual muscle movements.
The finding of this ‘direction tuning’ implied that the motor cortex cared more about kinematics than kinetics.
