数控编程g代码m代码，所有的！详细点！_百度知道(* Content-type: application/mathematica *)
(*** Wolfram Notebook File ***)
(* CreatedBy='Mathematica 6.0' *)
(*CacheID: 234*)
(* Internal cache information:
NotebookFileLineBreakTest
NotebookFileLineBreakTest
NotebookDataPosition[
NotebookDataLength[
NotebookOptionsPosition[
NotebookOutlinePosition[
CellTagsIndexPosition[
WindowFrame->Normal
ContainsDynamic->False*)
(* Beginning of Notebook Content *)
Notebook[{
Cell["\", "Subtitle",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Goals", "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Neural networks and the visual system", "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Retina-cortex pathway", "Subsubsection"],
Cell["\", "Text"],
Cell[TextData[{
StyleBox["Retinal-cortical pathway",
FontWeight->"Bold"],
": light->retina->thalamuscortex\nlight->(receptors->bipolar \
cells->ganglion cells)->(lateral geniculate cells)->V1 cells in layer 4C\n\
...also other neurons (e.g. in the retina, horizontal and amacrine cells)\n\
...also other pathways, e.g. to the superior colliculus"
}], "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Summary approximation analogous to generic model neuron:", \
"Subsubsection"],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
ImageSize->{427, 218.125},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell[TextData[{
"Despite several layers of synapses, many properties of ganglion cells, \
l.g.n. cells, and classes of V1 cells can be modeled using the generic model \
neuron, but with neural input replaced by light intensity. Where x",
Cell[BoxData[
SubscriptBox["", "i"], TraditionalForm]]],
Cell[BoxData[
SubscriptBox["L", "i"], TraditionalForm]]],
Cell[BoxData[
SubscriptBox["L", "i"], TraditionalForm]]],
"is the intensity of the light corresponding to location i. So firing rate \
}], "Text"],
Cell[BoxData[
SubscriptBox["R", "i"], "=",
RowBox[{"\[Sigma]",
RowBox[{"(",
UnderoverscriptBox["\[Sum]",
RowBox[{"j", "=", "1"}], "n"],
SubscriptBox["w", "ij"],
SubscriptBox["L", "j"]}]}], ")"}]}]}]], "Text"],
Cell["\", "Text"],
Cell["There is an analogous model for time:", "Text"],
Cell[BoxData[
SubscriptBox["R", "T"], "=",
RowBox[{"\[Sigma]",
RowBox[{"(",
UnderoverscriptBox["\[Sum]",
RowBox[{"t", "=", "1"}],
RowBox[{"T", "-", "1"}]],
SubscriptBox["w",
RowBox[{"T", ",", "t"}]],
SubscriptBox["L", "t"]}]}], ")"}]}]}]], "Text"]
}, Closed]],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[TextData[{
"Pathways from eye-to-cortex",
StyleBox["",
FontWeight->"Plain"]
}], "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Schematic view", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{468, 445.812},
ImageMargins->{{27, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}]
}, Closed]],
Cell[TextData[{
StyleBox["Retina",
FontWeight->"Bold"],
"\nThe primate retina has about 10^7 cones that send visual signals to the
optic nerve via about 10^6 ganglion cells. "
}], "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"The visual system in general responds well to contrasts over a middle \
range, but not as well to low or high spatial frequencies. The visual system \
is sometimes said to behave as a ",
StyleBox["bandpass spatial filter",
FontSlant->"Italic"],
". because it lets frequencies in a middle range of spatial frequencies pass \
through better than low or high frequencies. "
}], "Text"],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
ImageSize->{300., 69.8125}],
Cell["\", "Text"],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
ImageSize->{141., 112.25},
ImageMargins->{{70, 0}, {0, 13}}],
Cell["\", "Text"],
Cell[TextData[{
"Ganglion cells do spatio-temporal filtering of incoming image intensities. \
They are band-pass filters meaning that low and high frequencies of image \
intensity get suppressed, and only spatial and temporal changes in a middle \
band of frequencies gets passed on. Why?\n\t",
StyleBox["->Lateral inhibition as redundancy reduction is one important \
FontSlant->"Italic"]
}], "Text"],
Cell[TextData[{
StyleBox["Functions:",
FontWeight->"Bold"],
" image sampling, contrast coding, gain control, spatio-temporal filtering \
(high pass--encoding change important), efficient coding"
}], "Text"],
Cell[TextData[{
StyleBox["Optic chiasm",
FontWeight->"Bold"],
"\nThe optic nerves from the
two eyes meet at the optic chiasm where about \
half of the fibers cross over and
the other half remain on the same side of \
the underside of the brain. Before
synapsing in the lateral geniculate \
nucleus, about 20% of these fibers that
now comprise the optic tract branch \
off to the superior colliculus--a structure
involved with eye movements. The \
rest of the optic tract fibers
synapse on cells in the lateral geniculate \
nucleus. Cells in the lateral
geniculate nucleus send their axons in a \
bundle called the optic radiation
to layer IV (one of six layers) of primary \
visual cortex."
}], "Text"],
Cell[CellGroupData[{
Cell["Functions of the Chiasm and Lateral geniculate nucleus (LGN)", \
"Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"The optic chiasm routes neuronal information so that information
corresponding points on the left and right eyes can come together at
for binocular vision, and in particular stereo vision.
Typically animals \
with frontal vision have nearly complete cross-over, and animals with \
strongly lateralized eyes (e.g. fish, rabbits) have little or no cross-over. \
There is a correlation between eye position and whether an animal is \
primarily prey or predator.\nThe nervous system has gone to considerable \
to bring information from the two eyes together early on. Although \
there seems to be little if any binocular interaction between neurons in the \
LGN, the arrangment of the optic chiasm is the first step towards the \
eventual construction of a topographic cortical map with information from \
both eyes.\nIn fact, there is a general principle that becomes even more \
apparent when
one looks at maps that pervade cortical organization: \n",
StyleBox["Neural computations that require bringing information together \
often require close physical connectivity between neurons. ",
FontWeight->"Bold",
FontSlant->"Italic"],
"\nThis part of one solution to the \"binding problem\". There are other \
ways of bringing information together for a computation.\nLater on we will \
see some of the consequences of self-organizing principles that serve to \
minimize wiring length when we study ",
StyleBox["Kohonen networks and adaptive maps.",
FontWeight->"Bold"],
"\nThe neurons of lateral geniculate nucleus do more band-pass filtering,
and the cells are characterized by fairly symmetrical center-surround
organization like the ganglion cells. They show even less response to uniform \
illumination than ganglion cells. Despite the fact that neurons from the two \
eyes exist within the same nucleus, no binocular neurons are found in LGN.
We have to wait until cortex to see binocular neurons. Although the LGN is
often considered a relay station,
feedback from cortex suggests possible \
role of attention mechanisms (see Crick, 1984 for
a speculative neural \
network theory of
LGN and reticular function). Sillito et al.
(1994) have \
\"Feature-linked synchronization of thalamic relay cell firing induced \
by feedback from the visual cortex\". Nature, 369, N. 9, 479-482. An \
estimated ratio of 10 fibers returning to LGN for every one going forward \
raises a major computational question.\nSee Sherman and Guillery (2002) for a \
recent discussion of thalamo-cortical anatomy.\nThe superior colliculus has a \
key role is in the control of eye movements--a highly non-trivial
problem requiring coordination of head and eye movements in the context of
constantly changing environment."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Primary Visual Cortex", "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Cortex in general", "Subsection"],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Primary visual cortex: Large scale organization", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell[TextData[StyleBox["Overview ",
FontWeight->"Bold"]], "Subsubsection"],
Cell[TextData[{
"\tprimary visual cortex (striate cortex, V1 in primates, area 17 in cats)\n\
\t\tanatomical organization - ",
StyleBox["topographic, later lecture on adapative maps",
FontSlant->"Italic"],
"\n\t\tfunctional cell types: simple, complex, end-stopped\n\t\tmodel of \
simple cells\n\t\t\t",
StyleBox["generic feedforward neural network models",
FontSlant->"Italic"],
"\n\t\t\tother than sigmoid non-linearities?"
}], "Text",
Evaluatable->False,
TextAlignment->Left,
TextJustification->0,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Topographic map", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"The striate cortex is laid out as non-linear topographic map with 80%
area devoted to about 20% of visual field, reflecting the higher
acuity of foveal vision. Because of the cross-over at the optic chiasm, the
left visual field (right retina) maps to right hemisphere. (see Engel et al., \
1997 for fMRI measurments of ",
StyleBox["human cortical magnification",
FontSlant->"Italic"],
"). Nearby locations in the retina map to nearby locations in primary visual \
cortex--thus, V1 is said to have a ",
StyleBox["retinotopic map",
FontSlant->"Italic"],
".\nThere are a number of interesting features of this map, where locations \
in retinal polar coordinates are approximately mapped to rectangular \
coordinates on V1. Lecture 21 describes the properties of this map in more \
detail. \nWhy is visual information laid out topographically in many cortical \
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Layers", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Primary visual cortex: Neuron properties", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{427, 597.875},
ImageMargins->{{0, 0}, {0, 31}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell[TextData[{
"see Kandel and Schwartz, ",
ButtonBox["/exec/obidos/tg/detail/-//qid=\
/sr=1-1/ref=sr_1_1/103-6262?v=glance&s=books",
BaseStyle->"Hyperlink",
ButtonData->{
URL["/exec/obidos/tg/detail/-//qid=\
/sr=1-1/ref=sr_1_1/103-6262?v=glance&s=books"], None}]
}], "Text"],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Simple cells", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"There are two main types of cells. The ",
StyleBox["simple",
FontSlant->"Italic"],
cells are roughly linear except for
rectification, are spatially and \
temporally band-pass, and show spatial phase
sensitivity."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text"],
Cell[BoxData[{
RowBox[{"Plot", "[",
RowBox[{"Sin", "[", "x", "]"}], ",",
RowBox[{"{",
RowBox[{"x", ",",
RowBox[{"-", "12"}], ",", "12"}], "}"}]}],
"]"}], "\[IndentingNewLine]",
RowBox[{"Plot", "[",
RowBox[{"If", "[",
RowBox[{"Sin", "[", "x", "]"}], ">", "0"}], ",",
RowBox[{"Sin", "[", "x", "]"}], ",", "0"}], "]"}], ",",
RowBox[{"{",
RowBox[{"x", ",",
RowBox[{"-", "12"}], ",", "12"}], "}"}]}],
"]"}], "\[IndentingNewLine]",
RowBox[{"Plot", "[",
RowBox[{"If", "[",
RowBox[{"Sin", "[", "x", "]"}], ">", "0"}], ",",
RowBox[{"Sin", "[", "x", "]"}], ",",
RowBox[{"-",
RowBox[{"Sin", "[", "x", "]"}]}]}], "]"}], ",",
RowBox[{"{",
RowBox[{"x", ",",
RowBox[{"-", "12"}], ",", "12"}], "}"}]}], "]"}]}], "Input"],
Cell["\", "Text"],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{219, 91.75},
ImageMargins->{{140, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{366, 41.1875},
ImageMargins->{{0, 0}, {0, 5}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell["\", "Text"],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{111, 111},
ImageMargins->{{146, Inherited}, {Inherited, 11}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{236, 346}, {638.438, 528.438}} -> {-13.5, \
0..0463641}}],
Cell[TextData[{
"The squashing function is a half-wave rectification operation, ",
StyleBox["\[Sigma]",
FontFamily->"Symbol"],
sets negative values to zero, and is linear for positive values:"
}], "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{342, 196.938},
ImageMargins->{{87, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell[TextData[{
"You can see that the equation for R has the same form as the ",
StyleBox["generic neuron model",
FontWeight->"Bold"],
", except that the inputs are the physical stimulus values. The weights \
don't necessarily correspond to specific synapses, but are the net effect of \
a number of earlier layers.\nAnd as we saw at the begining of the course, a \
better model is obtained by replacing the straight sloping line with one that \
saturates at high values. This model is steady state. To include time domain \
dependencies requires the introduction of a band-pass temporal tuning \
characteristics.
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Complex cells", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"The second major class of neurons is that of ",
StyleBox["complex",
FontSlant->"Italic"],
cells. Like simple cells, complex cells are spatially and
temporally
band-pass, show orientation and motion direction selectivity, but are \
insensitive to
the phase of a stimulus such as a sine-wave grating. Rather \
than half-wave rectification, they show full-wave rectification. A model for \
complex cells would resemble the sum of the outputs of several
subunits \
positioned at several nearby spatial locations. Each subunit would resemble
simple cell with a linear spatial filter followed by a threshold \
non-linearity.\nIf true, the simplest neural-net like version of this model \
would correspond to two layers of weights, where the first set feed into \
simple cells, and the second set feed into complex cells. In actuality, \
complex cells may not be built out of simple cells, and as mentioned above, \
the generic connections model of simple cells collapses a number of neural \
layers to one effective layer. Another complication is that cells show a \
property called \"response normalization\" (see contrast normalization, \
below).\nOne way of obtaining the phase insensitivity would be to use \
subunits with cosine and sine phase receptive fields. We see below how a \
neural network can be built that can be used to detect edges--it
combines \
simple cell outputs into outputs similar to those of complex cells.\nSee \
Mechler, F., & Ringach, D. L. (2002) for a discussion of whether simple and \
complex cells make distinct classes.\nThe motion selectivity could be built \
in with appropriate inhibitory connections between subunits. Full-wave \
rectification could be built with subunit pairs that have excitatory and \
inhibitory receptive fields centers. "
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Contrast normalization", "Subsubsection"],
Cell["\", "Text"],
Cell[BoxData[
SubscriptBox["R", "i"], "=",
RowBox[{"\[Sigma]",
RowBox[{"(",
UnderoverscriptBox["\[Sum]",
RowBox[{"j", "=", "1"}], "n"],
SubscriptBox["w", "ij"],
SubscriptBox["L", "j"]}]}], ")"}], "/",
UnderscriptBox["\[Sum]",
RowBox[{"k", "\[Element]",
SubscriptBox["N", "i"]}]],
SubsuperscriptBox["R", "k", "2"]}]}]}]}]], "Text"],
Cell[BoxData[
RowBox[{"where", " ",
SubscriptBox["N", "i"], "is", " ", "a", " ", "neighborhood", " ", "of", " ",
"neuron", " ", "i"}]], "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["End-stopped cells", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"A third class of cells are the ",
StyleBox["end-stopped
FontSlant->"Italic"],
"(or \"hyper-complex\") cells
that have an optimal orientation for a bar or \
edge stimulus, but fire most actively if the bar or edge terminates within \
the receptive field, rather than extending beyond it. It has been suggested \
that these cells act as \"curvature\" detectors. (Dobbins, A., Zucker, S. W., \
& Cynader, M. S., 1987).\nThese cells are also thought to be important for \
detecting occluding surfaces and the perception of illusory contours. (Heiter \
et al., 1992).\nWhether or not these end-stopped cells should be considered a \
distinct functional class has been a matter of debate."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Functions of Primary Cortex", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell[TextData[StyleBox["Functions of primary visual cortex",
FontWeight->"Bold"]], "Subsubsection"],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Binocular vision and Stereo", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[TextData[{
StyleBox["Note:",
FontWeight->"Bold"],
" What is stereo disparity? Demo with thumb and forefinger."
}], "Text"],
Cell[TextData[{
"One possible algorithm for stereo vision is discussed in: Poggio, T. \
(1984). Vision by Man and Machine. ",
StyleBox["Scientific American",
FontVariations->{"Underline"->True}],
StyleBox["250",
FontVariations->{"Underline"->True}],
", 106-115. \nThis algorithm is related to ",
StyleBox["Hopfield networks",
FontWeight->"Bold"],
" that we will study later in this course.\nStereo vision has received a lot \
of attention in both computer and biological vision over the last 15 years. \
Later we will look at a neural network model of stereopsis."
}], "Text"],
Cell[CellGroupData[{
Cell["Motion", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\tmotion measurements--local in space and time", "Text"],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Spatio-temporal filtering", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["Overview", "Text"],
Cell[TextData[{
"\tSpatial frequency filtering, temporal filtering- ",
StyleBox["generic feedforward neural network models",
FontSlant->"Italic"],
"\t(Lecture 3)\n\t\tWhy spatial filtering?\n\t\t\tedge detection \n\t\t\t\
cortical basis set and economic representations - ",
StyleBox["self-organizing neural networks (Lecture 14)\n\t\t",
FontSlant->"Italic"],
"Why temporal filtering?\n\t\t\ttransient detection\n\t\t\teconomic \
representations",
StyleBox["\n\t",
FontSlant->"Italic"],
"Predictive coding? - A key theme in information processing by the brain"
}], "Text"],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[TextData[StyleBox["Spatial filtering",
FontWeight->"Bold"]], "Subsection",
Evaluatable->False,
AspectRatioFixed->True,
FontWeight->"Plain",
FontSlant->"Plain",
FontTracking->"Plain",
FontVariations->{"Outline"->False,
"Shadow"->False,
"Underline"->False}],
Cell["The convolution model for neural networks", "Subsubsection"],
Cell["\", "Text"],
Cell[TextData[{
Cell[BoxData[
SubscriptBox["r",
RowBox[{"k", ",", "l"}]], TraditionalForm]]],
" be the response (in spikes/sec) of a ganglion cell at x-y location (k,l). \
The average response, to a first approximation, is determined by the weighted \
sum of the inputs, ",
Cell[BoxData[
SubscriptBox["f",
RowBox[{"i", ",", "j"}]], TraditionalForm]]],
" at spatial location (i,j)"
}], "Text"],
Cell[TextData[{
Cell[BoxData[
SubscriptBox["r",
RowBox[{"k", ",", "l"}]], TraditionalForm]],
FontSize->18],
StyleBox["=",
FontSize->18],
Cell[BoxData[
UnderscriptBox["\[Sum]",
RowBox[{"i", ",", "j"}]],
SubscriptBox["w",
RowBox[{"k", ",",
RowBox[{"l", ";", " ", "i"}], ",", "j"}]],
SubscriptBox["f",
RowBox[{"i", ",", "j"}]]}]}], TraditionalForm]],
FontSize->18]
}], "Text"],
Cell["If we assume spatial homogeneity, and thus shift-invariance:", "Text"],
Cell[TextData[{
Cell[BoxData[
SubscriptBox["r",
RowBox[{"k", ",", "l"}]], TraditionalForm]],
FontSize->18],
StyleBox["=",
FontSize->18],
Cell[BoxData[
UnderscriptBox["\[Sum]",
RowBox[{"i", ",", "j"}]],
SubscriptBox["w",
RowBox[{"k", "-", "i"}], ";", " ",
RowBox[{"l", "-", "j"}]}]],
SubscriptBox["f",
RowBox[{"i", ",", "j"}]]}]}], TraditionalForm]],
FontSize->18],
". Or by suitable arrangement of rows and columns as matrix operation, ",
StyleBox["r = W.g",
FontWeight->"Bold"]
}], "Text"],
Cell["In the continuous case,by the convolution integral:", "Text"],
Cell[TextData[{
StyleBox["r(x,y) = w*g = ",
FontSize->18],
Cell[BoxData[
RowBox[{"\[Integral]",
RowBox[{"\[Integral]",
RowBox[{"w", "(",
RowBox[{"x", "-",
RowBox[{"x", "'"}]}], ",",
RowBox[{"y", "-",
RowBox[{"y", "'"}]}]}], ")"}],
RowBox[{"f", "(",
RowBox[{"x", "'"}], ",", "y"}], ")"}],
RowBox[{"\[DifferentialD]",
RowBox[{"x", "'"}]}],
RowBox[{"\[DifferentialD]",
RowBox[{"y", "'"}]}]}]}]}], TraditionalForm]],
FontSize->18]
}], "Text"],
Cell[TextData[{
"The fundamental computations in spatial filtering of input images are \
linear--image vectors are multiplied by a weight matrix. \nWhen it reached \
steady-state, the lateral inhibition network studied in Lecture 3 was \
essentially convolving the input with the weights. If you ever use a graphics \
package like Adobe Photoshop, you can easily convolve the image on the \
computer screen with any number of possible spatial filters (i.e. weight \
matrices). ",
StyleBox["Mathematica",
FontSlant->"Italic"],
" has a built-in function ",
StyleBox["ListConvolve[ ] ",
FontWeight->"Bold"],
"that accepts as arguments an input vector and a \"kernel\".",
"ut we need to know how to specify the \"kernel\", i.e. the weights for the \
convolution.\nSo the questions are: \n\tCan other neural systems be modeled \
as taking convolutions of their inputs? Yes. Simple cell responses are \
modeled as convolutions with a point-wise non-linearity.\n\tWhat is the \
structure of the effective weights for simple cell neurons in the visual \
cortex? Let's look at this more closely because it provides insight into \
issues of neural representation."
}], "Text"],
Cell[CellGroupData[{
Cell["Basis set for representing visual information", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
StyleBox["Psychophysics and physiology",
FontSlant->"Italic"],
"\nThe results of masking, adaptation, and other psychophysical studies of \
spatial and orientation frequency selectivity in human vision are \
surprisingly consistent.\nWhat is the form of the weights ",
Cell[BoxData[
SubscriptBox["w", "ij"], TraditionalForm]]],
Cell[BoxData[
SubscriptBox["w", "xy"], TraditionalForm]]],
StyleBox["A cortical basis set for images specifies the effective weights as \
a function of spatial position",
FontSlant->"Italic"],
"\n\tBoth the psychophysical and neurophysiological data could be accounted \
for, in part,
by assuming the visual system
performed a quasi-Fourier \
analysis of the image. One possible model assumes that the visual system \
computes the coefficients (or spectrum) of an image with respect to the \
following basis set, called a Gabor set (Daugman, 1988):"
}], "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{407, 56.375},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{299., 125.3125}],
Cell[TextData[{
"Here we've plotted a one-dimensional slice (e,g, ",
Cell[BoxData[
SubscriptBox["w", "x"], TraditionalForm]]],
") through a sine, and cosine Gabor function. In two dimensions (with the \
standard deviation , and the x and y spatial frequencies equal to 1), we can \
visualize the receptive field weights as follows."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell[" Visualizing the Gabor functions:", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->False],
Cell["\", "Input",
AspectRatioFixed->False],
Cell[CellGroupData[{
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{204., 44.25},
ImageMargins->{{72, 0}, {0, 0}}],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{233., 25.9375},
ImageMargins->{{63, 0}, {0, 0}}]
}, Closed]],
Cell[CellGroupData[{
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->False],
Cell["\", "Input",
AspectRatioFixed->False]
}, Closed]],
Cell[CellGroupData[{
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->False],
Cell[CellGroupData[{
Cell["\24, Mesh->False],
\t\t{i, 2, 3}, {j, 1, 3}] // Short
\>", "Input",
AspectRatioFixed->True],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{111, 111},
ImageMargins->{{146, Inherited}, {Inherited, 11}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{236, 346}, {341, 231}} -> {-13.07, 0.0463641, \
0.0463641}}],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{111, 111},
ImageMargins->{{146, Inherited}, {Inherited, 11}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{236, 346}, {471, 361}} -> {-13.4, 0.0463641, \
0.0463641}}],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{111, 111},
ImageMargins->{{146, Inherited}, {Inherited, 11}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{236, 346}, {601, 491}} -> {-13.7, 0.0463641, \
0.0463641}}]
}, Closed]]
}, Closed]]
}, Closed]],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Example: Visualizing receptive field weights as a basis set:", "Text"],
Cell[CellGroupData[{
Cell[GraphicsData["PostScript", "\"], "Graphics",
ImageSize->{288, 66.9375},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{89.875, 376.875}, {239.812, 173.875}} -> {-1.45709, \
2.58135}, {{98.5}, {238.188, 175.438}} -> \
{-8.536, 0..0662954}, {{167.438, 230.188}, {238.188, \
175.438}} -> {-13.36, 0..0662954}, {{236.5, 299.312}, \
{238.188, 175.438}} -> {-17.36, 0..0662954}, {{305.562, \
368.375}, {238.188, 175.438}} -> {-22.36, 0..0662954}}],
Cell[GraphicsData["PostScript", "\"], "Graphics",
ImageSize->{288, 66.9375},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{89.875, 376.875}, {314.75, 248.812}} -> {-1.45709, \
3.68135}, {{98.5}, {313.125, 250.375}} -> \
{-8.54, 0..0662954}, {{167.438, 230.188}, {313.125, \
250.375}} -> {-13.4, 0..0662954}, {{236.5, 299.312}, \
{313.125, 250.375}} -> {-17.4, 0..0662954}, {{305.562, \
368.375}, {313.125, 250.375}} -> {-22.4, 0..0662954}}],
Cell[GraphicsData["PostScript", "\"], "Graphics",
ImageSize->{288, 66.9375},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{89.875, 376.875}, {389.688, 323.75}} -> {-1.45709, \
4.8135}, {{98.5}, {388.062, 325.312}} -> \
{-8.54, 0..0662954}, {{167.438, 230.188}, {388.062, \
325.312}} -> {-13.4, 0..0662954}, {{236.5, 299.312}, \
{388.062, 325.312}} -> {-17.4, 0..0662954}, {{305.562, \
368.375}, {388.062, 325.312}} -> {-22.4, 0..0662954}}],
Cell[GraphicsData["PostScript", "\"], "Graphics",
ImageSize->{288, 66.9375},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{89.875, 376.875}, {464.625, 398.688}} -> {-1.45709, \
5.88135}, {{98.5}, {463, 400.25}} -> \
{-8.54, 0..0662954}, {{167.438, 230.188}, {463, \
400.25}} -> {-13.4, 0..0662954}, {{236.5, 299.312}, \
{463, 400.25}} -> {-17.4, 0..0662954}, {{305.562, \
368.375}, {463, 400.25}} -> {-22.4, 0..0662954}}]
}, Closed]]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["\", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell[CellGroupData[{
Cell["Edge Detection by Neural Networks", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["\", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text"],
Cell[CellGroupData[{
Cell["Define the filters", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Input",
AspectRatioFixed->False],
Cell[CellGroupData[{
Cell["\", "Input",
AspectRatioFixed->False],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{140, 86},
ImageMargins->{{37, Inherited}, {Inherited, Inherited}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{127, 266}, {190, 105}} -> {-3.264, 0.0165456, \
0.0191783}}]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Define the input stimulus: an ideal edge", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Plot[Sign[x],{x,-1,1},Axes->None]", "Input",
AspectRatioFixed->False],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{140, 86},
ImageMargins->{{37, Inherited}, {Inherited, Inherited}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{127, 266}, {356, 271}} -> {-3.038, 0.0152692, \
0.0247063}}]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Input",
AspectRatioFixed->False],
Cell[BoxData[
RowBox[{"Plot", "[",
RowBox[{"cr", "[", "x", "]"}], ",",
RowBox[{"{",
RowBox[{"x", ",",
RowBox[{"-", "1"}], ",", "1"}], "}"}], ",",
RowBox[{"PlotPoints", "\[Rule]", "10"}]}], "]"}]], "Input",
AspectRatioFixed->False],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{143, 88},
ImageMargins->{{122, Inherited}, {Inherited, Inherited}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{212, 354}, {571, 484}} -> {-4.602, 0.0162612, \
}, Closed]],
Cell[CellGroupData[{
Cell["Calculate the response of a bank of sine filters to the edge", \
"Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Input",
AspectRatioFixed->False],
Cell[BoxData[
RowBox[{"Plot", "[",
RowBox[{"sr", "[", "x", "]"}], ",",
RowBox[{"{",
RowBox[{"x", ",",
RowBox[{"-", "1"}], ",", "1"}], "}"}], ",",
RowBox[{"PlotPoints", "\[Rule]", "10"}]}], "]"}]], "Input",
AspectRatioFixed->False],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{156, 96},
ImageMargins->{{37, Inherited}, {Inherited, Inherited}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{127, 282}, {153, 58}} -> {-2.89268, -0..014026, \
}, Closed]],
Cell["\", "Subsubsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["Plot[cr[x]^2 + sr[x]^2,{x,-1,1},PlotPoints->10]", "Input",
AspectRatioFixed->False],
Cell[GraphicsData["PostScript", "\"], "Graphics",
Evaluatable->False,
AspectRatioFixed->True,
ImageSize->{177, 108},
ImageMargins->{{37, Inherited}, {Inherited, Inherited}},
ImageCache->GraphicsData["CompressedBitmap", "\"],
ImageRangeCache->{{{127, 303}, {339, 232}} -> {-2.974, 0.0136349, \
}, Closed]],
Cell[CellGroupData[{
Cell["\", "Text",
CellMargins->{{13, Inherited}, {Inherited, Inherited}},
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"Edges in the real world of natural images don't necessarily correspond to a \
sharp ideal edge, but can be blurry and noisy. Nevertheless they are ",
StyleBox["perceived",
FontSlant->"Italic"],
" to be sharp. \nMorrone and Burr went on to show that one could do the same \
operation with different sizes of filters (i.e. different values of ",
StyleBox["sigma",
FontWeight->"Bold"],
"), and each time the peak of the above operation for an ideal edge occurs \
at the edge transition. But even for blurry edges, the larger scale filters \
will still find a point in the transition region. Thus by adding up a whole \
set of neural outputs over a range of scales, one could detect an edge. \
Another way of viewing this network is one that detects ",
StyleBox["phase coherence",
FontWeight->"Bold"],
". Fourier theory shows that a step function can be built up of sine-waves \
of various frequencies whose zero crossings all line up with (say positive \
slope) at the edge transition."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[CellGroupData[{
Cell["\", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text"]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Economical representations by neurons in primary cortex", "Subsection",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"We might expect something like Fourier analysis of the image to result in \
efficient coding because of
the close relationship between Fourier rotations \
and Principal Components Analysis (e.g. Appendix A, Andrews, 1983). Fourier \
coefficients for natural images tend to be uncorrelated. Some work has been \
completed toward a functional explanation for
the orientation and spatial \
frequency tuning properties of
cortical receptive fields based on the \
statistics of natural
images (Field, 1994), but the story is far from \
complete. Barlow has argued that a decorrelated representation of sensory \
information is important for efficient learning (Barlow, 1990).",
StyleBox["\n",
FontSlant->"Italic"],
"There has been considerable work on the relationship between \
self-organizing models of visual cortex, and efficient coding of image \
information. For more on this, see:
Linsker, R. (1990), Barlow, H. B., & \
Foldiak, P. (1989), Olshausen and Field (1996), Simoncelli and Olshausn \
(2001). \nOlshausen and Field (1996) show that simple cell receptive field \
weights can emerge as a consequence of sparse coding. Imagine you don't know \
what the receptive field weights should be, and would like to write an \
algorithm to discover what the should be given simple principles. Olshausen \
and Field showed that if one enforces two constraints: 1) the image should be \
well-approximated by a weighted so 2) when representing \
natural images, the (squared) cooefficients (i.e. the projections of the \
image vector onto the basis vectors) should add up to a small number, then \
the receptive field weight structures emerge as a consequence.\nWe'll take a \
closer look at the topic of neural networks for efficient encoding in the \
next lecture (Lecture 14)."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell["Extra-striate cortical visual areas", "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell["\", "Text",
Evaluatable->False,
AspectRatioFixed->True],
Cell[GraphicsData["CompressedBitmap", "\"], "Graphics",
Evaluatable->False,
ImageSize->{392, 318.25},
ImageMargins->{{0, 0}, {0, 0}},
ImageRegion->{{0, 1}, {0, 1}}],
Cell[TextData[{
StyleBox["Spatial, action pathway ",
FontWeight->"Bold"],
"\n\tV1,MT,MST,LIP,...\nViewer-centered computations\n\"Where\" (vs. \
\"What\")\n",
StyleBox["(\"where\"",
FontWeight->"Bold"],
StyleBox["\"how\" ",
FontWeight->"Bold"],
StyleBox[" \"now\")",
FontWeight->"Bold"],
"\nSpatial computations,such as coordinate transformations for action"
}], "Text"],
Cell[TextData[{
StyleBox["Object perception, recognition pathway (\"what\")",
FontWeight->"Bold"],
"\n \tobject perception, recognition pathway\n \t\tV1, V2, V4, Posterior IT, \
Anterior IT, ... - ",
StyleBox["generic feedforward neural network models PLUS feedback",
FontSlant->"Italic"],
"\n \t\tInvariances required for recognition: \n \t\t\tphotometric: \
illumination level, direction, shadows\n \t\t\tgeometrical: translation, \
size, orientation in depth\n \t\t\tcategory-related: levels of abstraction\n \
\t\tBinding problem:\n \t\t\tgrandmother cells, distributed codes, sparse \
codes - \"binding by synchrony\""
}], "Text"],
Cell["\", "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["Modeling large-scale neural systems & systems analysis", "Section"],
Cell[TextData[{
"Much of the modeling of visual processing has been built on the tools that \
we've learned about. But there are many aspects of brain modeling that \
require additional tools and ways of thinking. \n",
StyleBox["Modeling information processing functions:",
FontWeight->"Bold"]
}], "Text"],
Cell[TextData[{
"\tFeedback\n\t\tInformation processing roles of feedback\n\t\t*Dynamical \
behavior\n\tTiming and sequences (e.g. speech, motor sequences)\n\tDynamical \
issues for real-time control, visuo-motor control\n\tLarge-scale \
architectures (e.g. Inter-area processing)\n\t*Handling uncertainty - \
Probabilistic models\n",
StyleBox["Measuring and characterizing neural systems",
FontWeight->"Bold"],
"\n\tLinear and non-linear systems analysis, statistical and stochastic \
processes analysis (time series),..."
}], "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell["References", "Section",
Evaluatable->False,
AspectRatioFixed->True],
Cell[TextData[{
"Adelson, E. H., & Bergen, J. R. (1985). Spatiotemporal Energy Models for \
the Perception of Motion. ",
StyleBox["Journal of the Optical Society of America",
FontVariations->{"Underline"->True}],
StyleBox["2",
FontVariations->{"Underline"->True}],
"((2)), 284-299.\nAdelson, E. H., Simoncelli, E., & Hingorani, R. (1987). \
Orthogonal Pyramid Transforms for Image Coding. Proc. SPIE - Visual \
Communication & Image Proc. II, Cambridge, MA. \nBarlow, H. B., & Foldiak, P. \
(1989). Adaptation and decorrelation in the cortex. In C. Miall, R. M. \
Durban, & G. J. Mitchison (Ed.), The Computing Neuron Addison-Wesley.\n\
Barlow, H. (1990). Conditions for versatile learning, Helmholtz's unconscious \
inference, and the task of perception. Vision Research, 30(11), .\n\
Boynton, G. M., Demb, J. B., Glover, G. H., & Heeger, D. J. (1999). Neuronal \
basis of contrast discrimination. ",
StyleBox["Vision Res, 39",
FontSlant->"Italic"],
"(2), 257-269.\nBullier, J. (2001). Integrated model of visual processing. \
StyleBox["Brain Res Brain Res Rev, 36",
FontSlant->"Italic"],
"(2-3), 96-107.\nCampbell, F. W., & Robson, J. R. (1968). Application of \
Fourier Analysis to the Visibility of Gratings. 197, 551-566.\nCarandini, M., \
Heeger, D. J., & Movshon, J. A. (1997). Linearity and normalization in simple \
cells of the macaque primary visual cortex. ",
StyleBox["J Neurosci, 17",
FontSlant->"Italic"],
"(21), .\nCrick, F. (1984). Function of the Thalamic Reticular \
Complex: The Searchlight Hypothesis. 81, .\nDaugman, J. G. (1988). \
An information-theoretic view of analog representation in striate cortex. In \
Computational Neuroscience Cambridge, Massachusetts: M.I.T. Press.\nDeValois, \
R., Albrecht, D. G., & Thorell, L. G. (1982). Spatial frequency selectivity \
of cells in macaque visual cortex. Vision Research, 22, 545-559), \nDobbins, \
A., Zucker, S. W., & Cynader, M. S. (1987). Endstopped neurons in the visual \
cortex as a substrate for calculating curvature. Nature, 329(6138), 438-441.\n\
Engel, S. A., Glover, G. H., & Wandell, B. A. (1997). Retinotopic \
organization in human visual cortex and the spatial precision of functional \
StyleBox["Cereb Cortex, 7",
FontSlant->"Italic"],
"(2), \nFang, F., Murray, S. O., Kersten, D. J., & He, S. (2005). \
Orientation-tuned fMRI adaptation in human visual cortex. J Neurophysiol.\n\
181-192.\nField, D. J. (1994). What is the goal of sensory coding? Nueral \
Computation, 6, 559-601.\nGeisler, W. S., & Albrecht, D. G. (1997). Visual \
cortex neurons in monkeys and cats: detection, discrimination, and \
identification. ",
StyleBox["Vis Neurosci, 14",
FontSlant->"Italic"],
"(5), 897-919.\nGeorgopoulos, A. P., Lurito, J. T., Petrides, M., Schwartz, \
A. B., & Massey, J. T. (1989). Mental Rotation of the Neuronal Population \
Vector. ",
StyleBox["Science, 243",
FontSlant->"Italic"],
", 234-236.\nHeeger, D. J. (1991). Nonlinear model of neural responses in \
cat visual cortex. In M. &. M. Landy A. (Ed.), Computational Models of Visual \
Processing (pp. 119-133). Cambridge, Massachusetts: M.I.T. Press.\nHeeger, D. \
J., Simoncelli, E. P., & Movshon, J. A. (1996). Computational Models of \
Cortical Visual Processing. Proc. National Academy of Science. 93, 623--627.\n\
Heitger, F., Rosenthaler, L., von der Heydt, R., Peterhans, E., & Kubler, O. \
(1992). Simulation of neural contour mechanisms: from simple to end-stopped \
StyleBox["Vision Res, 32",
FontSlant->"Italic"],
"(5), 963-981.\n",
StyleBox["Hubel, D. H., & Wiesel, T. N. (1968). ",
FontColor->RGBColor[0..]],
StyleBox["Receptive Fields and Functional Architecture of Monkey Striate \
FontVariations->{"Underline"->True},
FontColor->RGBColor[0..]],
StyleBox[" J. Physiol., London.pp. ",
FontColor->RGBColor[0..]],
"215-243.\nKoenderink, J. J., & van Doorn, A. J. (1990). Receptive field \
families. Biol. Cybern., 63, 291-297.\nLee, C., Rohrer, W. H., & Sparks, D. \
L. (1988). Population coding of saccadic eye movements by neurons in the \
superior colliculus. ",
StyleBox["Nature, 332",
FontSlant->"Italic"],
"(6162), 357-360.\nLennie, P. (2003). The cost of cortical computation. ",
StyleBox["Curr Biol, 13",
FontSlant->"Italic"],
"(6), 493-497.\nLinsker, R. (1990). Perceptual neural organization: some \
approaches based on network models and information theory. Annual Review of \
Neuroscience, 13, 257-281.\nLivingstone, M. S., & Hubel, D. H. (1984). \
Anatomy and Physiology of a Color System in the Primate Visual Cortex. 4(1), \
309-356; \nLivingstone, M. S., & Hubel, D. H. (1987). Psychophysical Evidence \
for Separate Channels for the Perception of Form, Color, Movement and Depth. \
The Journal of Neuroscience, 7(11), ).\n",
StyleBox["Mechler, F., & Ringach, D. L. (2002). On the classification of \
simple and complex cells. Vision Res, 42(8), .",
FontFamily->"Helvetica",
FontWeight->"Plain"],
"\nMumford, D. (1994). Neuronal architectures for pattern-theoretic \
problems. In C. Koch, & J. L. Davis (Ed.), ",
StyleBox["Large-Scale Neuronal Theories of the Brain",
FontVariations->{"Underline"->True}],
" (pp. 125-152). Cambridge, MA: MIT Press.\nMurray, S. O., Kersten, D., \
Olshausen, B. A., Schrater, P., & Woods, D. L. (2002). Shape perception \
reduces activity in human primary visual cortex. ",
StyleBox["Proc Natl Acad Sci U S A, 99",
FontSlant->"Italic"],
", .\nOlman, C., Ugurbil, K., Schrater, P., & Kersten, D. (2004). \
BOLD fMRI and psychophysical measurements of contrast response to broadband \
images. Vision Research, 44(7), 669-683.\nOlshausen, B. A., & Field, D. J. \
(1996). Emergence of simple-cell receptive field properties by learning a \
sparse code for natural images. ",
StyleBox["Nature",
FontVariations->{"Underline"->True}],
StyleBox["381",
FontVariations->{"Underline"->True}],
", 607-609.\nPoggio, G., F., & Poggio, T. ,1984. The Analysis of Stereopsis. \
Annual Review of Neuroscience, 7, 379-412). \nPoggio, T. (1984). Vision by \
Man and Machine. Scientific American, 250, 106-115.\nF. Rieke, D. Warland, R. \
de Ruyter van Steveninck, and W. B. (1996).",
StyleBox[" ",
FontVariations->{"CompatibilityType"->0}],
StyleBox["Spikes: Exploring the neural code",
FontVariations->{"CompatibilityType"->0},
FontColor->RGBColor[0...701961]],
StyleBox[" (",
FontVariations->{"CompatibilityType"->0}],
"MIT Press, Cambridge). \nPugh, M. C., Ringach, D. L., Shapley, R., & \
Shelley, M. J. (2000). Computational modeling of orientation tuning dynamics \
in monkey primary visual cortex. ",
StyleBox["J Comput Neurosci, 8",
FontSlant->"Italic"],
"(2), 143-159.\nSchwartz, E. L. (1980). Computational anatomy and functional \
architecture of striate cortex: a spatial mapping approach to perceptual \
coding. ",
StyleBox["Vision Res, 20",
FontSlant->"Italic"],
"(8), 645-669.\nShadlen, M. N., & Movshon, J. A. (1999). Synchrony unbound: \
a critical evaluation of the temporal binding hypothesis. ",
StyleBox["Neuron, 24",
FontSlant->"Italic"],
"(1), 67-77, 111-125.\nSherman, S. M., & Guillery, R. W. (2002). The role of \
the thalamus in the flow of information to the cortex. ",
StyleBox["Philos Trans R Soc Lond B Biol Sci, 357",
FontSlant->"Italic"],
"(1428), .\nSillito, A. M., Jones, H. E., Gerstein, G. L., & West, \
D. C. (1994). Feature-linked synchronization of thalamic relay cell firing \
induced by feedback from the visual cortex. Nature, 369, N. 9, 479-482.\n\
Simoncelli, E. P., & Heeger, D. J. (1998). A model of neuronal responses in \
visual area MT. ",
StyleBox["Vision Res",
FontVariations->{"Underline"->True}],
StyleBox["38",
FontVariations->{"Underline"->True}],
"(5), 743-61.\nSimoncelli, E. P., & Olshausen, B. A. (2001). Natural image \
statistics and neural representation. ",
StyleBox["Annu Rev Neurosci, 24",
FontSlant->"Italic"],
", .\nvon Melchner, L., Pallas, S. L., & Sur, M. (2000). Visual \
behaviour mediated by retinal projections directed to the auditory pathway. \
StyleBox["Nature, 404",
FontSlant->"Italic"],
"(6780), 871-876."
}], "Text",
Evaluatable->False,
AspectRatioFixed->True]
}, Closed]],
Cell["\", "Copyright"]
WindowToolbars->{"RulerBar", "EditBar"},
CellGrouping->Manual,
WindowSize->{905, 937},
WindowMargins->{{326, Automatic}, {Automatic, 4}},
DockedCells->(FrontEndExecute[{
FrontEnd`NotebookApply[
FrontEnd`InputNotebook[], #, Placeholder]}]& ),
PrintingCopies->1,
PrintingPageRange->{1, Automatic},
PrivateNotebookOptions->{"ColorPalette"->{RGBColor, 128}},
ShowCellLabel->True,
ShowCellTags->False,
RenderingOptions->{"ObjectDithering"->True,
"RasterDithering"->False},
CharacterEncoding->"MacintoshAutomaticEncoding",
FrontEndVersion->"6.0 for Mac OS X x86 (32-bit) (June 19, 2007)",
StyleDefinitions->"Classroom.nb"
(* End of Notebook Content *)
(* Internal cache information *)
(*CellTagsOutline
CellTagsIndex->{}
(*CellTagsIndex
CellTagsIndex->{}
(*NotebookFileOutline
Notebook[{
Cell[568, 21, 220, 9, 139, "Subtitle",
Evaluatable->False],
Cell[CellGroupData[{
Cell[813, 34, 70, 2, 51, "Section",
Evaluatable->False],
Cell[886, 38, 442, 9, 118, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[2, 2, 31, "Section",
Evaluatable->False],
Cell[CellGroupData[{
Cell[, 0, 33, "Subsubsection"],
Cell[0, 3, 22, "Text"],
Cell[5, 7, 103, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[, 1, 33, "Subsubsection"],
Cell[67, 81, 227, "Graphics",
Evaluatable->False],
Cell[, 571, 17, 37, "Text"],
Cell[, 281, 10, 51, "Text"],
Cell[, 579, 9, 79, "Text"],
Cell[, 53, 0, 22, "Text"],
Cell[, 335, 12, 51, "Text"]
}, Closed]],
Cell[, 522, 8, 52, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[, 145, 6, 31, "Section",
Evaluatable->False],
Cell[CellGroupData[{
Cell[, 82, 2, 41, "Subsection",
Evaluatable->False],
Cell[, 4, 454, "Graphics",
Evaluatable->False]
}, Closed]],
Cell[5, 240, 7, 49, "Text",
Evaluatable->False],
Cell[5, 404, 8, 37, "Text"],
Cell[5, , "Graphics",
Evaluatable->False],
Cell[5, 299, 5, 37, "Text"],
Cell[5, 4, "Graphics",
Evaluatable->False],
Cell[6, 367, 6, 37, "Text"],
Cell[6, 412, 8, 64, "Text"],
Cell[6, 210, 5, 22, "Text"],
Cell[6, 714, 12, 94, "Text"],
Cell[CellGroupData[{
Cell[6, 130, 3, 41, "Subsection",
Evaluatable->False],
Cell[6, 8, "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[6, 86, 2, 31, "Section",
Evaluatable->False],
Cell[CellGroupData[{
Cell[6, 39, 0, 41, "Subsection"],
Cell[6, 623, 11, 157, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[6, 115, 2, 41, "Subsection",
Evaluatable->False],
Cell[CellGroupData[{
Cell[6, 75, 1, 33, "Subsubsection"],
Cell[6, 537, 14, 157, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[6, 86, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[6, 973, 19, 121, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[7, 77, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[7, 352, 7, 37, "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[7, 108, 2, 41, "Subsection",
Evaluatable->False],
Cell[7, 873, 17, 229, "Text",
Evaluatable->False],
Cell[CellGroupData[{
Cell[7, 143, 4, 33, "Subsubsection",
Evaluatable->False],
Cell[7, 907, 15, 109, "Text",
Evaluatable->False],
Cell[7, 3, 637, "Graphics",
Evaluatable->False],
Cell[31, 390, 8, 22, "Text"],
Cell[41, 376, 6, 37, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[52, 83, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[56, 297, 8, 22, "Text",
Evaluatable->False],
Cell[66, 143, 3, 22, "Text"],
Cell[71, 898, 30, 75, "Input"],
Cell[03, 113, 3, 22, "Text"],
Cell[08, 0, "Graphics",
Evaluatable->False],
Cell[33, 251, 6, 37, "Text",
Evaluatable->False],
Cell[41, , "Graphics",
Evaluatable->False],
Cell[75, 153, 3, 22, "Text"],
Cell[80, 1, 130, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False],
Cell[41, 259, 7, 23, "Text",
Evaluatable->False],
Cell[50, 5, "Graphics",
Evaluatable->False],
Cell[78, 679, 13, 79, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[96, 84, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[00, 5, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[34, 47, 0, 33, "Subsubsection"],
Cell[36, 249, 4, 37, "Text"],
Cell[42, 462, 16, 52, "Text"],
Cell[60, 150, 3, 22, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[68, 88, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[72, 768, 14, 106, "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[92, 95, 2, 41, "Subsection",
Evaluatable->False],
Cell[CellGroupData[{
Cell[98, 100, 1, 33, "Subsubsection"],
Cell[01, 183, 4, 22, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[10, 98, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[14, 696, 12, 67, "Text",
Evaluatable->False]
}, Closed]],
Cell[29, 131, 4, 22, "Text"],
Cell[35, 595, 14, 76, "Text"],
Cell[CellGroupData[{
Cell[53, 77, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[57, 62, 0, 22, "Text"],
Cell[59, 629, 11, 67, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[75, 96, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[79, 24, 0, 22, "Text"],
Cell[81, 585, 13, 211, "Text"],
Cell[96, 459, 8, 52, "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[10, 269, 9, 41, "Subsection",
Evaluatable->False],
Cell[21, 66, 0, 33, "Subsubsection"],
Cell[23, 5, "Text"],
Cell[45, 443, 14, 42, "Text"],
Cell[61, 481, 20, 32, "Text"],
Cell[83, 76, 0, 22, "Text"],
Cell[85, 618, 24, 32, "Text"],
Cell[11, 67, 0, 22, "Text"],
Cell[13, 600, 22, 35, "Text"],
Cell[37, 5, "Text"],
Cell[CellGroupData[{
Cell[63, 116, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[67, 3, "Text",
Evaluatable->False],
Cell[93, , "Graphics",
Evaluatable->False],
Cell[37, 804, 14, 109, "Text",
Evaluatable->False],
Cell[53, 4, "Graphics",
Evaluatable->False],
Cell[99, 407, 10, 37, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[14, 105, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[18, 207, 6, 83, "Input"],
Cell[CellGroupData[{
Cell[28, 166, 5, 33, "Subsubsection",
Evaluatable->False],
Cell[35, , "Graphics",
Evaluatable->False],
Cell[65, , "Graphics",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[93, 152, 4, 33, "Subsubsection",
Evaluatable->False],
Cell[99, 104, 4, 57, "Input"]
}, Closed]],
Cell[CellGroupData[{
Cell[08, 240, 6, 55, "Subsubsection",
Evaluatable->False],
Cell[CellGroupData[{
Cell[18, 252, 9, 122, "Input"],
Cell[29, 1, 130, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False],
Cell[90, 1, 130, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False],
Cell[41, 1, 130, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]]
}, Closed]]
}, Closed]],
Cell[05, 2, "Text",
Evaluatable->False],
Cell[CellGroupData[{
Cell[28, 76, 0, 22, "Text"],
Cell[CellGroupData[{
Cell[32, 4, 75, 3, "GraphicsData", "PostScript", \
"Graphics"],
Cell[38, 4, 75, 3, "GraphicsData", "PostScript", \
"Graphics"],
Cell[06, 4, 75, 3, "GraphicsData", "PostScript", \
"Graphics"],
Cell[32, 4, 75, 3, "GraphicsData", "PostScript", \
"Graphics"]
}, Closed]]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[05, 140, 4, 41, "Subsection",
Evaluatable->False],
Cell[11, 981, 15, 109, "Text",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[31, 101, 2, 41, "Subsection",
Evaluatable->False],
Cell[35, 612, 10, 67, "Text",
Evaluatable->False],
Cell[CellGroupData[{
Cell[49, 159, 5, 41, "Subsection",
Evaluatable->False],
Cell[56, 87, 2, 22, "Text"],
Cell[CellGroupData[{
Cell[62, 89, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[66, 186, 6, 83, "Input"],
Cell[CellGroupData[{
Cell[76, 87, 4, 57, "Input"],
Cell[82, , 94, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[82, 111, 2, 33, "Subsubsection",
Evaluatable->False],
Cell[CellGroupData[{
Cell[88, 76, 1, 44, "Input"],
Cell[91, , 1324, 72, "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[80, 141, 4, 33, "Subsubsection",
Evaluatable->False],
Cell[86, 124, 4, 57, "Input"],
Cell[92, 266, 8, 44, "Input"],
Cell[02, , 96, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[41, 133, 3, 33, "Subsubsection",
Evaluatable->False],
Cell[46, 122, 4, 57, "Input"],
Cell[52, 266, 8, 44, "Input"],
Cell[62, , 104, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]],
Cell[79, 141, 4, 33, "Subsubsection",
Evaluatable->False],
Cell[CellGroupData[{
Cell[87, 90, 1, 40, "Input"],
Cell[90, , 116, , "GraphicsData", "PostScript", \
"Graphics",
Evaluatable->False]
}, Closed]],
Cell[CellGroupData[{
Cell[020, 566, 12, 145, "Text",
Evaluatable->False],
Cell[034, , "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[059, , "Text",
Evaluatable->False],
Cell[CellGroupData[{
Cell[080, 149, 4, 41, "Subsection",
Evaluatable->False],
Cell[086, 192, 4, 22, "Text"]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[096, 123, 2, 41, "Subsection",
Evaluatable->False],
Cell[100, 8, "Text",
Evaluatable->False]
}, Closed]]
}, Closed]],
Cell[CellGroupData[{
Cell[135, 100, 2, 31, "Section",
Evaluatable->False],
Cell[139, 520, 11, 79, "Text",
Evaluatable->False],
Cell[152, 5, 327, "Graphics",
Evaluatable->False],
Cell[050, 400, 14, 157, "Text"],
Cell[066, 643, 12, 238, "Text"],
Cell[080, 151, 3, 22, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[088, 73, 0, 31, "Section"],
Cell[090, 309, 6, 64, "Text"],
Cell[098, 540, 10, 238, "Text"]
}, Closed]],
Cell[CellGroupData[{
Cell[113, 75, 2, 31, "Section",
Evaluatable->False],
Cell[117, , 1318, "Text",
Evaluatable->False]
}, Closed]],
Cell[284, 167, 3, 54, "Copyright"]
(* End of internal cache information *)}