Related
I have two list x and y. I combine them and receive list of point {x,y}. Using ListLinPlot I build a graphic. My question is - Can I find the function using points? I mean find formula/function f(x) if I know only points.
For an example :
{{2,5},{3,7},{7,15},{9,19}}
So the answer will be:
F(x)=2x+1
It was easy example, but my graphic is not a liner function. The blue line is my function. I changed picture that better describe the problem
See https://reference.wolfram.com/language/ref/FindFormula.html
data = {{2, 5}, {3, 7}, {7, 15}, {9, 19}};
fit = FindFormula[data, x]
1. + 2. x
I have a very large array of numbers in the form of a third order tensor.I want to find the highest of all the values in that tensor. How can I do it in mathematica? The context is that a reaction is carried out by varying temperature pressure and vessel volume. I want to find the optimum combination of the three to maximize the product. Each element of the tensor represents a value of the product produced corresponding to a specific combination of temperature pressure and volume.
Given some matrix, tensor, or basically any list-of-lists of real numbers, you can simply use the Max function to determine the maximum value and then Position to say where it is. Assuming your data isn't enormous (requiring some conservative/careful approach to save time/memory), this should be fine.
For example, here is a random list of lists of of lists of reals:
data = Table[RandomReal[],
{i, 1, RandomInteger[{4, 8}]},
{j, 1, RandomInteger[{4, 8}]},
{k, 1, RandomInteger[{4, 8}]}
];
You can just do:
m = Max[data]
Position[data, m]
This will tell you the position of the maximum value. If you did random integers instead, you could have ties, in which case you might have repeats:
data = Table[RandomInteger[{1, 10}],
{i, 1, RandomInteger[{4, 8}]},
{j, 1, RandomInteger[{4, 8}]},
{k, 1, RandomInteger[{4, 8}]}
];
m = Max[data]
Position[data, m]
Table[RandomInteger[100, 3], 3]
Prepend[Ordering[%[[First[Ordering[Reverse#*Sort /# %, -1]]]], -1],
First[Ordering[Reverse#*Sort /# %, -1]]]
% stands for the tensor to sort, in this case it's a random tensor generated from Table[RandomInteger[100, 3], 3]
This gives the position and value in one shot.
(m = RandomReal[{-1, 1}, {4, 3, 2}]) // MatrixForm
First#MaximalBy[
Flatten[MapIndexed[ {##} &, #, {-1}], ArrayDepth[#] - 1],
First] &#m
{0.903213, {3, 2, 2}}
Here is an alternate that will work with ragged lists:
Module[{h},
First#MaximalBy[List ### Flatten[MapIndexed[h### &, #, {-1}]],
First]] &#{{1, 2, 3}, {4, 5, {2, 3}}}
{5, {2, 2}}
I have a file which, when I import into Mathematica, looks like this:
{{1,1,n1},{1,2,n2},{1,3,n3},{2,1,n4},{2,2,n5},{2,3,n6}} where n1...n6 are some numbers that I want to import as a matrix that looks like :
The first number in each block specifies the row and the second the column, but they are not a part of the matrix. Only the third number in each block is part of the matrix. How can I do that?
If
data = {{1, 1, n1}, {1, 2, n2}, {1, 3, n3}, {2, 1, n4}, {2, 2, n5}, {2, 3, n6}};
you can simply do
mat = Partition[data[[All, 3]], 3, 3]
There are a couple of interpretations of this question that I can think of.
If your data is in a regular format and you wish to read it in a memory efficient manner I recommend looking closely at ReadList and related functionality as I already directed you toward and the Partition function that the other answer illustrates.
I shall instead focus on the idea that the data is not in an entirely regular form in that the given row and column indexes are necessary to describe the positions of the data in the array. For that the most natural method is to use SparseArray at it accepts data in the form of position and value Rule pairs:
data = {{1, 1, n1}, {1, 2, n2}, {1, 3, n3}, {2, 1, n4}, {2, 2, n5}, {2, 3, n6}};
array = SparseArray[{#, #2} -> #3 & ### data];
array // MatrixForm
The function Normal can be used to convert the SparseArray into a regular list-of-lists array as needed:
Normal # array
{{n1, n2, n3}, {n4, n5, n6}}
Also there is a StackExchange site dedicated to Mathematica that I encourage you to explore.
Suppose I have two very large lists {a1, a2, …} and {b1, b2, …} where all ai and bj are large sparse arrays. For the sake of memory efficiency I store each list as one comprehensive sparse array.
Now I would like to compute some function f on all possible pairs of ai and bj where each result f[ai, bj] is a sparse array again. All these sparse arrays have the same dimensions, by the way.
While
Flatten[Outer[f, {a1, a2, ...}, {b1, b2, ...}, 1], 1]
returns the desired result (in principle) it appears to consume excessive amounts of memory. Not the least because the return value is a list of sparse arrays whereas one comprehensive sparse array turns out much more efficient in my cases of interest.
Is there an efficient alternative to the above use of Outer?
More specific example:
{SparseArray[{{1, 1, 1, 1} -> 1, {2, 2, 2, 2} -> 1}],
SparseArray[{{1, 1, 1, 2} -> 1, {2, 2, 2, 1} -> 1}],
SparseArray[{{1, 1, 2, 1} -> 1, {2, 2, 1, 2} -> 1}],
SparseArray[{{1, 1, 2, 2} -> -1, {2, 2, 1, 1} -> 1}],
SparseArray[{{1, 2, 1, 1} -> 1, {2, 1, 2, 2} -> 1}],
SparseArray[{{1, 2, 1, 2} -> 1, {2, 1, 2, 1} -> 1}],
SparseArray[{{1, 2, 2, 1} -> -1, {2, 1, 1, 2} -> 1}],
SparseArray[{{1, 2, 2, 2} -> 1, {2, 1, 1, 1} -> 1}]};
ByteCount[%]
list = SparseArray[%%]
ByteCount[%]
Flatten[Outer[Dot, list, list, 1], 1];
ByteCount[%]
list1x2 = SparseArray[%%]
ByteCount[%]
Flatten[Outer[Dot, list1x2, list, 1], 1];
ByteCount[%]
list1x3 = SparseArray[%%]
ByteCount[%]
etc. Not only are the raw intermediate results of Outer (lists of sparse arrays) extremely inefficient, Outer seems to consume way too much memory during the computation itself, too.
I will propose a solution which is rather complex but allows one to only use about twice as much memory during the computation as is needed to store the final result as a SparseArray. The price to pay for this will be a much slower execution.
The code
Sparse array construction / deconstruction API
Here is the code. First, a slightly modified (to address higher-dimensional sparse arrays) sparse array construction - deconstruction API, taken from this answer:
ClearAll[spart, getIC, getJR, getSparseData, getDefaultElement,
makeSparseArray];
HoldPattern[spart[SparseArray[s___], p_]] := {s}[[p]];
getIC[s_SparseArray] := spart[s, 4][[2, 1]];
getJR[s_SparseArray] := spart[s, 4][[2, 2]];
getSparseData[s_SparseArray] := spart[s, 4][[3]];
getDefaultElement[s_SparseArray] := spart[s, 3];
makeSparseArray[dims_List, jc_List, ir_List, data_List, defElem_: 0] :=
SparseArray ## {Automatic, dims, defElem, {1, {jc, ir}, data}};
Iterators
The following functions produce iterators. Iterators are a good way to encapsulate the iteration process.
ClearAll[makeTwoListIterator];
makeTwoListIterator[fname_Symbol, a_List, b_List] :=
With[{indices = Flatten[Outer[List, a, b, 1], 1]},
With[{len = Length[indices]},
Module[{i = 0},
ClearAll[fname];
fname[] := With[{ind = ++i}, indices[[ind]] /; ind <= len];
fname[] := Null;
fname[n_] :=
With[{ind = i + 1}, i += n;
indices[[ind ;; Min[len, ind + n - 1]]] /; ind <= len];
fname[n_] := Null;
]]];
Note that I could have implemented the above function more memory - efficiently and not use Outer in it, but for our purposes this won't be the major concern.
Here is a more specialized version, which produces interators for pairs of 2-dimensional indices.
ClearAll[make2DIndexInterator];
make2DIndexInterator[fname_Symbol, i : {iStart_, iEnd_}, j : {jStart_, jEnd_}] :=
makeTwoListIterator[fname, Range ## i, Range ## j];
make2DIndexInterator[fname_Symbol, ilen_Integer, jlen_Integer] :=
make2DIndexInterator[fname, {1, ilen}, {1, jlen}];
Here is how this works:
In[14]:=
makeTwoListIterator[next,{a,b,c},{d,e}];
next[]
next[]
next[]
Out[15]= {a,d}
Out[16]= {a,e}
Out[17]= {b,d}
We can also use this to get batch results:
In[18]:=
makeTwoListIterator[next,{a,b,c},{d,e}];
next[2]
next[2]
Out[19]= {{a,d},{a,e}}
Out[20]= {{b,d},{b,e}}
, and we will be using this second form.
SparseArray - building function
This function will build a SparseArray object iteratively, by getting chunks of data (also in SparseArray form) and gluing them together. It is basically code used in this answer, packaged into a function. It accepts the code piece used to produce the next chunk of data, wrapped in Hold (I could alternatively make it HoldAll)
Clear[accumulateSparseArray];
accumulateSparseArray[Hold[getDataChunkCode_]] :=
Module[{start, ic, jr, sparseData, dims, dataChunk},
start = getDataChunkCode;
ic = getIC[start];
jr = getJR[start];
sparseData = getSparseData[start];
dims = Dimensions[start];
While[True, dataChunk = getDataChunkCode;
If[dataChunk === {}, Break[]];
ic = Join[ic, Rest#getIC[dataChunk] + Last#ic];
jr = Join[jr, getJR[dataChunk]];
sparseData = Join[sparseData, getSparseData[dataChunk]];
dims[[1]] += First[Dimensions[dataChunk]];
];
makeSparseArray[dims, ic, jr, sparseData]];
Putting it all together
This function is the main one, putting it all together:
ClearAll[sparseArrayOuter];
sparseArrayOuter[f_, a_SparseArray, b_SparseArray, chunkSize_: 100] :=
Module[{next, wrapperF, getDataChunkCode},
make2DIndexInterator[next, Length#a, Length#b];
wrapperF[x_List, y_List] := SparseArray[f ### Transpose[{x, y}]];
getDataChunkCode :=
With[{inds = next[chunkSize]},
If[inds === Null, Return[{}]];
wrapperF[a[[#]] & /# inds[[All, 1]], b[[#]] & /# inds[[All, -1]]]
];
accumulateSparseArray[Hold[getDataChunkCode]]
];
Here, we first produce the iterator which will give us on demand portions of index pair list, used to extract the elements (also SparseArrays). Note that we will generally extract more than one pair of elements from two large input SparseArray-s at a time, to speed up the code. How many pairs we process at once is governed by the optional chunkSize parameter, which defaults to 100. We then construct the code to process these elements and put the result back into SparseArray, where we use an auxiliary function wrapperF. The use of iterators wasn't absolutely necessary (could use Reap-Sow instead, as with other answers), but allowed me to decouple the logic of iteration from the logic of generic accumulation of sparse arrays.
Benchmarks
First we prepare large sparse arrays and test our functionality:
In[49]:=
arr = {SparseArray[{{1,1,1,1}->1,{2,2,2,2}->1}],SparseArray[{{1,1,1,2}->1,{2,2,2,1}->1}],
SparseArray[{{1,1,2,1}->1,{2,2,1,2}->1}],SparseArray[{{1,1,2,2}->-1,{2,2,1,1}->1}],
SparseArray[{{1,2,1,1}->1,{2,1,2,2}->1}],SparseArray[{{1,2,1,2}->1,{2,1,2,1}->1}]};
In[50]:= list=SparseArray[arr]
Out[50]= SparseArray[<12>,{6,2,2,2,2}]
In[51]:= larger = sparseArrayOuter[Dot,list,list]
Out[51]= SparseArray[<72>,{36,2,2,2,2,2,2}]
In[52]:= (large= sparseArrayOuter[Dot,larger,larger])//Timing
Out[52]= {0.047,SparseArray[<2592>,{1296,2,2,2,2,2,2,2,2,2,2}]}
In[53]:= SparseArray[Flatten[Outer[Dot,larger,larger,1],1]]==large
Out[53]= True
In[54]:= MaxMemoryUsed[]
Out[54]= 21347336
Now we do the power tests
In[55]:= (huge= sparseArrayOuter[Dot,large,large,2000])//Timing
Out[55]= {114.344,SparseArray[<3359232>,{1679616,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2}]}
In[56]:= MaxMemoryUsed[]
Out[56]= 536941120
In[57]:= ByteCount[huge]
Out[57]= 262021120
In[58]:= (huge1 = Flatten[Outer[Dot,large,large,1],1]);//Timing
Out[58]= {8.687,Null}
In[59]:= MaxMemoryUsed[]
Out[59]= 2527281392
For this particular example, the suggested method is 5 times more memory-efficient than the direct use of Outer, but about 15 times slower. I had to tweak the chunksize parameter (default is 100, but for the above I used 2000, to get the optimal speed / memory use combination). My method only used as a peak value twice as much memory as needed to store the final result. The degree of memory-savings as compared to Outer- based method will depend on the sparse arrays in question.
If lst1 and lst2 are your lists,
Reap[
Do[Sow[f[#1[[i]], #2[[j]]]],
{i, 1, Length##1},
{j, 1, Length##2}
] &[lst1, lst2];
] // Last // Last
does the job and may be more memory-efficient. On the other hand, maybe not. Nasser is right, an explicit example would be useful.
EDIT: Using Nasser's randomly-generated arrays, and for len=200, MaxMemoryUsed[] indicates that this form needs 170MB while the Outer form in the question takes 435MB.
Using your example list data, I believe that you will find the ability to Append to a SparseArray quite helpful.
acc = SparseArray[{}, {1, 2, 2, 2, 2, 2, 2}]
Do[AppendTo[acc, i.j], {i, list}, {j, list}]
Rest[acc]
I need Rest to drop the first zero-filled tensor in the result. The second argument of the seed SparseArray must be the dimensions of each of your elements with a prefixed 1. You may need to explicitly specify a background for the seed SparseArray to optimize performance.
I'm looking for a fast implementation for the following, I'll call it Position2D for lack of a better term:
Position2D[ matrix, sub_matrix ]
which finds the locations of sub_matrix inside matrix and returns the upper left and lower right row/column of a match.
For example, this:
Position2D[{
{0, 1, 2, 3},
{1, 2, 3, 4},
{2, 3, 4, 5},
{3, 4, 5, 6}
}, {
{2, 3},
{3, 4}
}]
should return this:
{
{{1, 3}, {2, 4}},
{{2, 2}, {3, 3}},
{{3, 1}, {4, 2}}
}
It should be fast enough to work quickly on 3000x2000 matrices with 100x100 sub-matrices. For simplicity, it is enough to only consider integer matrices.
Algorithm
The following code is based on an efficient custom position function to find positions of (possibly overlapping) integer sequences in a large integer list. The main idea is that we can first try to eficiently find the positions where the first row of the sub-matrix is in the Flatten-ed large matrix, and then filter those, extracting full sub-matrices and comparing to the sub-matrix of interest. This will be efficient for most cases except very pathological ones (those, for which this procedure would generate a huge number of potential position candidates, while the true number of entries of the sub-matrix would be much smaller. But such cases seem rather unlikely generally, and then, further improvements to this simple scheme can be made).
For large matrices, the proposed solution will be about 15-25 times faster than the solution of #Szabolcs when a compiled version of sequence positions function is used, and 3-5 times faster for the top-level implementation of sequence positions - finding function. The actual speedup depends on matrix sizes, it is more for larger matrices. The code and benchmarks are below.
Code
A generally efficient function for finding positions of a sub-list (sequence)
These helper functions are due to Norbert Pozar and taken from this Mathgroup thread. They are used to efficiently find starting positions of an integer sequence in a larger list (see the mentioned post for details).
Clear[seqPos];
fdz[v_] := Rest#DeleteDuplicates#Prepend[v, 0];
seqPos[list_List, seq_List] :=
Fold[
fdz[#1 (1 - Unitize[list[[#1]] - #2])] + 1 &,
fdz[Range[Length[list] - Length[seq] + 1] *
(1 - Unitize[list[[;; -Length[seq]]] - seq[[1]]])] + 1,
Rest#seq
] - Length[seq];
Example of use:
In[71]:= seqPos[{1,2,3,2,3,2,3,4},{2,3,2}]
Out[71]= {2,4}
A faster position-finding function for integers
However fast seqPos might be, it is still the major bottleneck in my solution. Here is a compiled-to-C version of this, which gives another 5x performance boost to my code:
seqposC =
Compile[{{list, _Integer, 1}, {seq, _Integer, 1}},
Module[{i = 1, j = 1, res = Table[0, {Length[list]}], ctr = 0},
For[i = 1, i <= Length[list], i++,
If[list[[i]] == seq[[1]],
While[j < Length[seq] && i + j <= Length[list] &&
list[[i + j]] == seq[[j + 1]],
j++
];
If[j == Length[seq], res[[++ctr]] = i];
j = 1;
]
];
Take[res, ctr]
], CompilationTarget -> "C", RuntimeOptions -> "Speed"]
Example of use:
In[72]:= seqposC[{1, 2, 3, 2, 3, 2, 3, 4}, {2, 3, 2}]
Out[72]= {2, 4}
The benchmarks below have been redone with this function (also the code for main function is slightly modified )
Main function
This is the main function. It finds positions of the first row in a matrix, and then filters them, extracting the sub-matrices at these positions and testing against the full sub-matrix of interest:
Clear[Position2D];
Position2D[m_, what_,seqposF_:Automatic] :=
Module[{posFlat, pos2D,sp = If[seqposF === Automatic,seqposC,seqposF]},
With[{dm = Dimensions[m], dwr = Reverse#Dimensions[what]},
posFlat = sp[Flatten#m, First#what];
pos2D =
Pick[Transpose[#], Total[Clip[Reverse#dm - # - dwr + 2, {0, 1}]],2] &#
{Mod[posFlat, #, 1], IntegerPart[posFlat/#] + 1} &#Last[dm];
Transpose[{#, Transpose[Transpose[#] + dwr - 1]}] &#
Select[pos2D,
m[[Last## ;; Last## + Last#dwr - 1,
First## ;; First## + First#dwr - 1]] == what &
]
]
];
For integer lists, the faster compiled subsequence position-finding function seqposC can be used (this is a default). For generic lists, one can supply e.g. seqPos, as a third argument.
How it works
We will use a simple example to dissect the code and explain its inner workings. This defines our test matrix and sub-matrix:
m = {{0, 1, 2, 3}, {1, 2, 3, 4}, {2, 3, 4, 5}};
what = {{2, 3}, {3, 4}};
This computes the dimensions of the above (it is more convenient to work with reversed dimensions for a sub-matrix):
In[78]:=
dm=Dimensions[m]
dwr=Reverse#Dimensions[what]
Out[78]= {3,4}
Out[79]= {2,2}
This finds a list of starting positions of the first row ({2,3} here) in the Flattened main matrix. These positions are at the same time "flat" candidate positions of the top left corner of the sub-matrix:
In[77]:= posFlat = seqPos[Flatten#m, First#what]
Out[77]= {3, 6, 9}
This will reconstruct the 2D "candidate" positions of the top left corner of a sub-matrix in a full matrix, using the dimensions of the main matrix:
In[83]:= posInterm = Transpose#{Mod[posFlat,#,1],IntegerPart[posFlat/#]+1}&#Last[dm]
Out[83]= {{3,1},{2,2},{1,3}}
We can then try using Select to filter them out, extracting the full sub-matrix and comparing to what, but we'll run into a problem here:
In[84]:=
Select[posInterm,
m[[Last##;;Last##+Last#dwr-1,First##;;First##+First#dwr-1]]==what&]
During evaluation of In[84]:= Part::take: Cannot take positions 3 through 4
in {{0,1,2,3},{1,2,3,4},{2,3,4,5}}. >>
Out[84]= {{3,1},{2,2}}
Apart from the error message, the result is correct. The error message itself is due to the fact that for the last position ({1,3}) in the list, the bottom right corner of the sub-matrix will be outside the main matrix. We could of course use Quiet to simply ignore the error messages, but that's a bad style. So, we will first filter those cases out, and this is what the line Pick[Transpose[#], Total[Clip[Reverse#dm - # - dwr + 2, {0, 1}]], 2] &# is for. Specifically, consider
In[90]:=
Reverse#dm - # - dwr + 2 &#{Mod[posFlat, #, 1],IntegerPart[posFlat/#] + 1} &#Last[dm]
Out[90]= {{1,2,3},{2,1,0}}
The coordinates of the top left corners should stay within a difference of dimensions of matrix and a sub-matrix. The above sub-lists were made of x and y coordiantes of top - left corners. I added 2 to make all valid results strictly positive. We have to pick only coordiantes at those positions in Transpose#{Mod[posFlat, #, 1], IntegerPart[posFlat/#] + 1} &#Last[dm] ( which is posInterm), at which both sub-lists above have strictly positive numbers. I used Total[Clip[...,{0,1}]] to recast it into picking only at those positions at which this second list has 2 (Clip converts all positive integers to 1, and Total sums numbers in 2 sublists. The only way to get 2 is when numbers in both sublists are positive).
So, we have:
In[92]:=
pos2D=Pick[Transpose[#],Total[Clip[Reverse#dm-#-dwr+2,{0,1}]],2]&#
{Mod[posFlat,#,1],IntegerPart[posFlat/#]+1}&#Last[dm]
Out[92]= {{3,1},{2,2}}
After the list of 2D positions has been filtered, so that no structurally invalid positions are present, we can use Select to extract the full sub-matrices and test against the sub-matrix of interest:
In[93]:=
finalPos =
Select[pos2D,m[[Last##;;Last##+Last#dwr-1,First##;;First##+First#dwr-1]]==what&]
Out[93]= {{3,1},{2,2}}
In this case, both positions are genuine. The final thing to do is to reconstruct the positions of the bottom - right corners of the submatrix and add them to the top-left corner positions. This is done by this line:
In[94]:= Transpose[{#,Transpose[Transpose[#]+dwr-1]}]&#finalPos
Out[94]= {{{3,1},{4,2}},{{2,2},{3,3}}}
I could have used Map, but for a large list of positions, the above code would be more efficient.
Example and benchmarks
The original example:
In[216]:= Position2D[{{0,1,2,3},{1,2,3,4},{2,3,4,5},{3,4,5,6}},{{2,3},{3,4}}]
Out[216]= {{{3,1},{4,2}},{{2,2},{3,3}},{{1,3},{2,4}}}
Note that my index conventions are reversed w.r.t. #Szabolcs' solution.
Benchmarks for large matrices and sub-matrices
Here is a power test:
nmat = 1000;
(* generate a large random matrix and a sub-matrix *)
largeTestMat = RandomInteger[100, {2000, 3000}];
what = RandomInteger[10, {100, 100}];
(* generate upper left random positions where to insert the submatrix *)
rposx = RandomInteger[{1,Last#Dimensions[largeTestMat] - Last#Dimensions[what] + 1}, nmat];
rposy = RandomInteger[{1,First#Dimensions[largeTestMat] - First#Dimensions[what] + 1},nmat];
(* insert the submatrix nmat times *)
With[{dwr = Reverse#Dimensions[what]},
Do[largeTestMat[[Last#p ;; Last#p + Last#dwr - 1,
First#p ;; First#p + First#dwr - 1]] = what,
{p,Transpose[{rposx, rposy}]}]]
Now, we test:
In[358]:= (ps1 = position2D[largeTestMat,what])//Short//Timing
Out[358]= {1.39,{{{1,2461},{100,2560}},<<151>>,{{1900,42},{1999,141}}}}
In[359]:= (ps2 = Position2D[largeTestMat,what])//Short//Timing
Out[359]= {0.062,{{{2461,1},{2560,100}},<<151>>,{{42,1900},{141,1999}}}}
(the actual number of sub-matrices is smaller than the number we try to generate, since many of them overlap and "destroy" the previously inserted ones - this is so because the sub-matrix size is a sizable fraction of the matrix size in our benchmark).
To compare, we should reverse the x-y indices in one of the solutions (level 3), and sort both lists, since positions may have been obtained in different order:
In[360]:= Sort#ps1===Sort[Reverse[ps2,{3}]]
Out[360]= True
I do not exclude a possibility that further optimizations are possible.
This is my implementation:
position2D[m_, k_] :=
Module[{di, dj, extractSubmatrix, pos},
{di, dj} = Dimensions[k] - 1;
extractSubmatrix[{i_, j_}] := m[[i ;; i + di, j ;; j + dj]];
pos = Position[ListCorrelate[k, m], ListCorrelate[k, k][[1, 1]]];
pos = Select[pos, extractSubmatrix[#] == k &];
{#, # + {di, dj}} & /# pos
]
It uses ListCorrelate to get a list of potential positions, then filters those that actually match. It's probably faster on packed real matrices.
As per Leonid's suggestion here's my solution. I know it isn't very efficient (it's about 600 times slower than Leonid's when I timed it) but it's very short, rememberable, and a nice illustration of a rarely used function, PartitionMap. It's from the Developer package, so it needs a Needs["Developer`"] call first.
Given that, Position2D can be defined as:
Position2D[m_, k_] := Position[PartitionMap[k == # &, m, Dimensions[k], {1, 1}], True]
This only gives the upper-left coordinates. I feel the lower-right coordinates are really redundant, since the dimensions of the sub-matrix are known, but if the need arises one can add those to the output by prepending {#, Dimensions[k] + # - {1, 1}} & /# to the above definition.
How about something like
Position2D[bigMat_?MatrixQ, smallMat_?MatrixQ] :=
Module[{pos, sdim = Dimensions[smallMat] - 1},
pos = Position[bigMat, smallMat[[1, 1]]];
Quiet[Select[pos, (MatchQ[
bigMat[[Sequence##Thread[Span[#, # + sdim]]]], smallMat] &)],
Part::take]]
which will return the top left-hand positions of the submatrices.
Example:
Position2D[{{0, 1, 2, 3}, {1, 2, 3, 4}, {2, 3, 4, 5}, {3, 5, 5, 6}},
{{2, 3}, {3, _}}]
(* Returns: {{1, 3}, {2, 2}, {3, 1}} *)
And to search a 1000x1000 matrix, it takes about 2 seconds on my old machine
SeedRandom[1]
big = RandomInteger[{0, 10}, {1000, 1000}];
Position2D[big, {{1, 1, _}, {1, 1, 1}}] // Timing
(* {1.88012, {{155, 91}, {295, 709}, {685, 661},
{818, 568}, {924, 45}, {981, 613}}} *)