DelayedTensor 1.6.0
Authors: Koki Tsuyuzaki [aut, cre]
Last modified: 2023-04-25 14:41:07.038824
Compiled: Tue Apr 25 16:52:49 2023
einsumeinsum is an easy and intuitive way to write tensor operations.
It was originally introduced by
Numpy1 https://numpy.org/doc/stable/reference/generated/numpy.einsum.html
package of Python but similar tools have been implemented in other languages
(e.g. R, Julia) inspired by Numpy.
In this vignette, we will use CRAN einsum package first.
einsum is named after
Einstein summation2 https://en.wikipedia.org/wiki/Einstein_notation
introduced by Albert Einstein,
which is a notational convention that implies summation over
a set of indexed terms in a formula.
Here, we consider a simple example of einsum; matrix multiplication.
If we naively implement the matrix multiplication,
the calculation would look like the following in a for loop.
A <- matrix(runif(3*4), nrow=3, ncol=4)
B <- matrix(runif(4*5), nrow=4, ncol=5)
C <- matrix(0, nrow=3, ncol=5)
I <- nrow(A)
J <- ncol(A)
K <- ncol(B)
for(i in 1:I){
for(j in 1:J){
for(k in 1:K){
C[i,k] = C[i,k] + A[i,j] * B[j,k]
}
}
}Therefore, any programming language can implement this. However, when analyzing tensor data, such operations tend to be more complicated and increase the possibility of causing bugs because the order of tensors is larger or more tensors are handled simultaneously. In addition, several programming languages, especially R, are known to significantly slow down the speed of computation if the code is written in for loop.
Obviously, in the case of the R language, it should be executed using the built-in matrix multiplication function (%*%) prepared by the R, as shown below.
C <- A %*% BHowever, more complex operations than matrix multiplication are not always provided by programming languages as standard.
einsum is a function that solves such a problem.
To put it simply, einsum is a wrapper for the for loop above.
Like the Einstein summation, it omits many notations such as for,
array size (e.g. I, J, and K), brackets (e.g. {}, (), and []),
and even addition operator (+) and
extracts the array subscripts (e.g. i, j, and k)
to concisely express the tensor operation as follows.
suppressPackageStartupMessages(library("einsum"))
C <- einsum('ij,jk->ik', A, B)DelayedTensorCRAN einsum is easy to use because the syntax is almost
the same as that of Numpy‘s einsum,
except that it prohibits the implicit modes that do not use’->’.
It is extremely fast because the internal calculation
is actually performed by C++.
When the input tensor is huge, however,
it is not scalable because it assumes that the input is R’s standard array.
Using einsum of DelayedTensor,
we can augment the CRAN einsum’s functionality;
in DelayedTensor,
the input DelayedArray objects are divided into
multiple block tensors and the CRAN einsum
is incremently applied in the block processing.
A surprisingly large number of tensor operations can be handled
uniformly in einsum.
In more detail, einsum is capable of performing any tensor operation
that can be described by a combination of the following
three operations3 https://ajcr.net/Basic-guide-to-einsum/.
Some typical operations are introduced below. Here we use the arrays and DelayedArray objects below.
suppressPackageStartupMessages(library("DelayedTensor"))
suppressPackageStartupMessages(library("DelayedArray"))
arrA <- array(runif(3), dim=c(3))
arrB <- array(runif(3*3), dim=c(3,3))
arrC <- array(runif(3*4), dim=c(3,4))
arrD <- array(runif(3*3*3), dim=c(3,3,3))
arrE <- array(runif(3*4*5), dim=c(3,4,5))
darrA <- DelayedArray(arrA)
darrB <- DelayedArray(arrB)
darrC <- DelayedArray(arrC)
darrD <- DelayedArray(arrD)
darrE <- DelayedArray(arrE)If the same subscript is written on both sides of ->,
einsum will simply output the object without any calculation.
einsum::einsum('i->i', arrA)## [1] 0.4435179 0.2612365 0.1691130DelayedTensor::einsum('i->i', darrA)## <3> array of class DelayedArray and type "double":
## [1] [2] [3]
## 0.4435179 0.2612365 0.1691130einsum::einsum('ij->ij', arrC)## [,1] [,2] [,3] [,4]
## [1,] 0.2306521 0.5597838 0.6111953 0.2317208
## [2,] 0.5596185 0.8827677 0.6571940 0.5501385
## [3,] 0.6256027 0.7088709 0.6379488 0.3687220DelayedTensor::einsum('ij->ij', darrC)## <3 x 4> matrix of class DelayedArray and type "double":
## [,1] [,2] [,3] [,4]
## [1,] 0.2306521 0.5597838 0.6111953 0.2317208
## [2,] 0.5596185 0.8827677 0.6571940 0.5501385
## [3,] 0.6256027 0.7088709 0.6379488 0.3687220einsum::einsum('ijk->ijk', arrE)## , , 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.3941131 0.9341609 0.1456407 0.3450411
## [2,] 0.1669718 0.5141893 0.3516905 0.7500159
## [3,] 0.3027837 0.7690442 0.9640399 0.8668125
##
## , , 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.5636724 0.3606369 0.31591087 0.1523008
## [2,] 0.2043268 0.7859820 0.86592401 0.6572030
## [3,] 0.3297063 0.5692190 0.07859974 0.2063811
##
## , , 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.6726905 0.4249505 0.7551751 0.6907411
## [2,] 0.5275158 0.7882609 0.2572507 0.9604261
## [3,] 0.1213238 0.4637394 0.6904535 0.8812318
##
## , , 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.4671706 0.3672498 0.3245687 0.7914497
## [2,] 0.2554462 0.6546781 0.2081990 0.7536588
## [3,] 0.4345269 0.3591480 0.9887159 0.5133280
##
## , , 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.6756489 0.7937759 0.03827615 0.6392523
## [2,] 0.1504246 0.9076869 0.86114266 0.6103908
## [3,] 0.9857722 0.7501000 0.21644762 0.7059945DelayedTensor::einsum('ijk->ijk', darrE)## <3 x 4 x 5> array of class DelayedArray and type "double":
## ,,1
## [,1] [,2] [,3] [,4]
## [1,] 0.3941131 0.9341609 0.1456407 0.3450411
## [2,] 0.1669718 0.5141893 0.3516905 0.7500159
## [3,] 0.3027837 0.7690442 0.9640399 0.8668125
##
## ,,2
## [,1] [,2] [,3] [,4]
## [1,] 0.56367245 0.36063694 0.31591087 0.15230084
## [2,] 0.20432677 0.78598197 0.86592401 0.65720303
## [3,] 0.32970635 0.56921899 0.07859974 0.20638113
##
## ,,3
## [,1] [,2] [,3] [,4]
## [1,] 0.6726905 0.4249505 0.7551751 0.6907411
## [2,] 0.5275158 0.7882609 0.2572507 0.9604261
## [3,] 0.1213238 0.4637394 0.6904535 0.8812318
##
## ,,4
## [,1] [,2] [,3] [,4]
## [1,] 0.4671706 0.3672498 0.3245687 0.7914497
## [2,] 0.2554462 0.6546781 0.2081990 0.7536588
## [3,] 0.4345269 0.3591480 0.9887159 0.5133280
##
## ,,5
## [,1] [,2] [,3] [,4]
## [1,] 0.67564890 0.79377592 0.03827615 0.63925226
## [2,] 0.15042456 0.90768685 0.86114266 0.61039081
## [3,] 0.98577219 0.75010002 0.21644762 0.70599451We can also extract the diagonal elements as follows.
einsum::einsum('ii->i', arrB)## [1] 0.848674765 0.005582589 0.622618967DelayedTensor::einsum('ii->i', darrB)## <3> array of class HDF5Array and type "double":
## [1] [2] [3]
## 0.848674765 0.005582589 0.622618967einsum::einsum('iii->i', arrD)## [1] 0.8355310 0.5134344 0.8110374DelayedTensor::einsum('iii->i', darrD)## <3> array of class HDF5Array and type "double":
## [1] [2] [3]
## 0.8355310 0.5134344 0.8110374By using multiple arrays or DelayedArray objects as input and writing “,” on the right side of ->, multiplication will be performed.
Hadamard Product can also be implemented in einsum,
multiplying by the product of each element.
einsum::einsum('i,i->i', arrA, arrA)## [1] 0.19670811 0.06824449 0.02859921DelayedTensor::einsum('i,i->i', darrA, darrA)## <3> array of class HDF5Array and type "double":
## [1] [2] [3]
## 0.19670811 0.06824449 0.02859921einsum::einsum('ij,ij->ij', arrC, arrC)## [,1] [,2] [,3] [,4]
## [1,] 0.05320037 0.3133579 0.3735597 0.05369451
## [2,] 0.31317281 0.7792788 0.4319039 0.30265232
## [3,] 0.39137871 0.5024979 0.4069787 0.13595593DelayedTensor::einsum('ij,ij->ij', darrC, darrC)## <3 x 4> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3] [,4]
## [1,] 0.05320037 0.31335786 0.37355968 0.05369451
## [2,] 0.31317281 0.77927880 0.43190392 0.30265232
## [3,] 0.39137871 0.50249792 0.40697873 0.13595593einsum::einsum('ijk,ijk->ijk', arrE, arrE)## , , 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.15532515 0.8726566 0.02121121 0.1190534
## [2,] 0.02787959 0.2643906 0.12368617 0.5625239
## [3,] 0.09167795 0.5914290 0.92937299 0.7513639
##
## , , 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.31772663 0.1300590 0.099799681 0.02319555
## [2,] 0.04174943 0.6177677 0.749824387 0.43191583
## [3,] 0.10870628 0.3240103 0.006177919 0.04259317
##
## , , 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.45251246 0.1805830 0.57028940 0.4771233
## [2,] 0.27827287 0.6213553 0.06617791 0.9224183
## [3,] 0.01471947 0.2150542 0.47672598 0.7765695
##
## , , 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.21824841 0.1348724 0.10534484 0.6263926
## [2,] 0.06525275 0.4286035 0.04334682 0.5680015
## [3,] 0.18881364 0.1289873 0.97755922 0.2635057
##
## , , 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.45650143 0.6300802 0.001465064 0.4086435
## [2,] 0.02262755 0.8238954 0.741566679 0.3725769
## [3,] 0.97174682 0.5626500 0.046849572 0.4984282DelayedTensor::einsum('ijk,ijk->ijk', darrE, darrE)## <3 x 4 x 5> array of class HDF5Array and type "double":
## ,,1
## [,1] [,2] [,3] [,4]
## [1,] 0.15532515 0.87265664 0.02121121 0.11905337
## [2,] 0.02787959 0.26439064 0.12368617 0.56252392
## [3,] 0.09167795 0.59142900 0.92937299 0.75136385
##
## ,,2
## [,1] [,2] [,3] [,4]
## [1,] 0.317726628 0.130059001 0.099799681 0.023195547
## [2,] 0.041749430 0.617767662 0.749824387 0.431915827
## [3,] 0.108706276 0.324010259 0.006177919 0.042593170
##
## ,,3
## [,1] [,2] [,3] [,4]
## [1,] 0.45251246 0.18058297 0.57028940 0.47712325
## [2,] 0.27827287 0.62135530 0.06617791 0.92241825
## [3,] 0.01471947 0.21505423 0.47672598 0.77656948
##
## ,,4
## [,1] [,2] [,3] [,4]
## [1,] 0.21824841 0.13487242 0.10534484 0.62639261
## [2,] 0.06525275 0.42860345 0.04334682 0.56800153
## [3,] 0.18881364 0.12898731 0.97755922 0.26350566
##
## ,,5
## [,1] [,2] [,3] [,4]
## [1,] 0.456501431 0.630080214 0.001465064 0.408643452
## [2,] 0.022627549 0.823895418 0.741566679 0.372576941
## [3,] 0.971746820 0.562650046 0.046849572 0.498428248The outer product can also be implemented in einsum,
in which the subscripts in the input array are all different,
and all of them are kept.
einsum::einsum('i,j->ij', arrA, arrA)## [,1] [,2] [,3]
## [1,] 0.19670811 0.11586305 0.07500464
## [2,] 0.11586305 0.06824449 0.04417848
## [3,] 0.07500464 0.04417848 0.02859921DelayedTensor::einsum('i,j->ij', darrA, darrA)## <3 x 3> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3]
## [1,] 0.19670811 0.11586305 0.07500464
## [2,] 0.11586305 0.06824449 0.04417848
## [3,] 0.07500464 0.04417848 0.02859921einsum::einsum('ij,klm->ijklm', arrC, arrE)## , , 1, 1, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.0909030 0.2206181 0.2408801 0.09132419
## [2,] 0.2205530 0.3479103 0.2590088 0.21681679
## [3,] 0.2465582 0.2793753 0.2514240 0.14531819
##
## , , 2, 1, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.03851239 0.09346811 0.1020524 0.03869084
## [2,] 0.09344051 0.14739733 0.1097329 0.09185762
## [3,] 0.10445802 0.11836146 0.1065195 0.06156619
##
## , , 3, 1, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.06983767 0.1694934 0.1850600 0.07016126
## [2,] 0.16944333 0.2672876 0.1989876 0.16657294
## [3,] 0.18942227 0.2146345 0.1931605 0.11164301
##
## , , 1, 2, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2154661 0.5229281 0.5709548 0.2164645
## [2,] 0.5227737 0.8246471 0.6139249 0.5139179
## [3,] 0.5844136 0.6621995 0.5959469 0.3444457
##
## , , 2, 2, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1185988 0.2878348 0.3142701 0.1191483
## [2,] 0.2877498 0.4539097 0.3379221 0.2828753
## [3,] 0.3216782 0.3644938 0.3280265 0.1895929
##
## , , 3, 2, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1773816 0.4304985 0.4700362 0.1782035
## [2,] 0.4303713 0.6788874 0.5054112 0.4230808
## [3,] 0.4811161 0.5451530 0.4906109 0.2835635
##
## , , 1, 3, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.03359232 0.08152728 0.08901489 0.03374797
## [2,] 0.08150321 0.12856688 0.09571417 0.08012254
## [3,] 0.09111320 0.10324043 0.09291130 0.05370092
##
## , , 2, 3, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.08111812 0.1968706 0.2149515 0.08149398
## [2,] 0.19681247 0.3104610 0.2311288 0.19347844
## [3,] 0.22001849 0.2493031 0.2243605 0.12967602
##
## , , 3, 3, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2223578 0.5396539 0.5892167 0.2233881
## [2,] 0.5394945 0.8510233 0.6335612 0.5303554
## [3,] 0.6031060 0.6833798 0.6150082 0.3554628
##
## , , 1, 4, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.07958444 0.1931484 0.2108875 0.07995319
## [2,] 0.19309137 0.3045911 0.2267589 0.18982039
## [3,] 0.21585864 0.2445896 0.2201186 0.12722426
##
## , , 2, 4, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1729927 0.4198467 0.4584062 0.1737943
## [2,] 0.4197228 0.6620899 0.4929060 0.4126126
## [3,] 0.4692120 0.5316645 0.4784718 0.2765474
##
## , , 3, 4, 1
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1999321 0.4852275 0.5297917 0.2008584
## [2,] 0.4850843 0.7651940 0.5696639 0.4768669
## [3,] 0.5422802 0.6144581 0.5529820 0.3196128
##
## , , 1, 1, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1300122 0.3155347 0.3445139 0.1306146
## [2,] 0.3154415 0.4975918 0.3704421 0.3100979
## [3,] 0.3526350 0.3995710 0.3595942 0.2078384
##
## , , 2, 1, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.04712839 0.1143788 0.1248836 0.04734675
## [2,] 0.11434503 0.1803731 0.1342823 0.11240802
## [3,] 0.12782738 0.1448413 0.1303500 0.07533978
##
## , , 3, 1, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.07604745 0.1845643 0.2015150 0.0763998
## [2,] 0.18450976 0.2910541 0.2166810 0.1813841
## [3,] 0.20626517 0.2337192 0.2103358 0.1215700
##
## , , 1, 2, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.08318165 0.2018787 0.2204196 0.08356706
## [2,] 0.20181909 0.3183586 0.2370084 0.19840025
## [3,] 0.22561543 0.2556450 0.2300679 0.13297478
##
## , , 2, 2, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1812884 0.4399799 0.4803885 0.1821283
## [2,] 0.4398500 0.6938395 0.5165426 0.4323989
## [3,] 0.4917124 0.5571597 0.5014163 0.2898089
##
## , , 3, 2, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1312915 0.3186395 0.3479040 0.1318999
## [2,] 0.3185455 0.5024881 0.3740873 0.3131493
## [3,] 0.3561049 0.4035028 0.3631326 0.2098836
##
## , , 1, 3, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.07286549 0.1768418 0.1930832 0.07320311
## [2,] 0.17678955 0.2788759 0.2076147 0.17379472
## [3,] 0.19763469 0.2239400 0.2015350 0.11648330
##
## , , 2, 3, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1997271 0.4847302 0.5292487 0.2006526
## [2,] 0.4845871 0.7644097 0.5690800 0.4763781
## [3,] 0.5417244 0.6138283 0.5524152 0.3192853
##
## , , 3, 3, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.01812919 0.04399886 0.04803979 0.01821319
## [2,] 0.04398587 0.06938531 0.05165528 0.04324074
## [3,] 0.04917221 0.05571707 0.05014261 0.02898146
##
## , , 1, 4, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.03512850 0.08525554 0.09308556 0.03529127
## [2,] 0.08523036 0.13444626 0.10009120 0.08378655
## [3,] 0.09527981 0.10796163 0.09716015 0.05615668
##
## , , 2, 4, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1515852 0.3678916 0.4016794 0.1522876
## [2,] 0.3677829 0.5801576 0.4319099 0.3615527
## [3,] 0.4111480 0.4658721 0.4192619 0.2423252
##
## , , 3, 4, 2
##
## [,1] [,2] [,3] [,4]
## [1,] 0.04760223 0.1155288 0.1261392 0.04782279
## [2,] 0.11549469 0.1821866 0.1356324 0.11353820
## [3,] 0.12911259 0.1462976 0.1316606 0.07609727
##
## , , 1, 1, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1551574 0.3765612 0.4111452 0.1558763
## [2,] 0.3764500 0.5938294 0.4420881 0.3700729
## [3,] 0.4208370 0.4768507 0.4291421 0.2480358
##
## , , 2, 1, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1216726 0.2952948 0.3224151 0.1222363
## [2,] 0.2952076 0.4656739 0.3466802 0.2902067
## [3,] 0.3300153 0.3739406 0.3365281 0.1945067
##
## , , 3, 1, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.02798359 0.06791510 0.07415255 0.02811325
## [2,] 0.06789505 0.10710075 0.07973328 0.06674490
## [3,] 0.07590051 0.08600292 0.07739839 0.04473477
##
## , , 1, 2, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.09801572 0.2378804 0.2597278 0.09846986
## [2,] 0.23781017 0.3751326 0.2792749 0.23378164
## [3,] 0.26585020 0.3012351 0.2710967 0.15668863
##
## , , 2, 2, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1818140 0.4412557 0.4817814 0.1826564
## [2,] 0.4411254 0.6958513 0.5180403 0.4336527
## [3,] 0.4931382 0.5587752 0.5028702 0.2906492
##
## , , 3, 2, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1069624 0.2595938 0.2834353 0.1074580
## [2,] 0.2595171 0.4093742 0.3047667 0.2551209
## [3,] 0.2901166 0.3287314 0.2958420 0.1709909
##
## , , 1, 3, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1741827 0.4227347 0.4615595 0.1749897
## [2,] 0.4226099 0.6666442 0.4962965 0.4154509
## [3,] 0.4724396 0.5353216 0.4817631 0.2784497
##
## , , 2, 3, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.0593354 0.1440048 0.1572304 0.05961032
## [2,] 0.1439622 0.2270926 0.1690636 0.14152349
## [3,] 0.1609367 0.1823575 0.1641128 0.09485399
##
## , , 3, 3, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1592545 0.3865046 0.4220019 0.1599924
## [2,] 0.3863905 0.6095100 0.4537619 0.3798450
## [3,] 0.4319495 0.4894424 0.4404740 0.2545854
##
## , , 1, 4, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1593208 0.3866656 0.4221777 0.1600590
## [2,] 0.3865515 0.6097639 0.4539509 0.3800032
## [3,] 0.4321295 0.4896462 0.4406575 0.2546915
##
## , , 2, 4, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2215242 0.5376309 0.5870079 0.2225507
## [2,] 0.5374722 0.8478331 0.6311862 0.5283673
## [3,] 0.6008451 0.6808181 0.6127027 0.3541303
##
## , , 3, 4, 3
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2032579 0.4932992 0.5386047 0.2041997
## [2,] 0.4931536 0.7779230 0.5791402 0.4847995
## [3,] 0.5513010 0.6246796 0.5621808 0.3249296
##
## , , 1, 1, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1077539 0.2615145 0.2855325 0.1082531
## [2,] 0.2614373 0.4124032 0.3070217 0.2570085
## [3,] 0.2922632 0.3311637 0.2980310 0.1722561
##
## , , 2, 1, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.05891919 0.1429946 0.1561275 0.05919218
## [2,] 0.14295240 0.2254996 0.1678777 0.14053077
## [3,] 0.15980782 0.1810784 0.1629616 0.09418864
##
## , , 3, 1, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1002245 0.2432411 0.2655808 0.1006889
## [2,] 0.2431693 0.3835863 0.2855685 0.2390500
## [3,] 0.2718412 0.3080235 0.2772059 0.1602196
##
## , , 1, 2, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.08470692 0.2055805 0.2244614 0.0850994
## [2,] 0.20551977 0.3241963 0.2413544 0.2020382
## [3,] 0.22975247 0.2603327 0.2342866 0.1354131
##
## , , 2, 2, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1510029 0.3664782 0.4001362 0.1517025
## [2,] 0.3663700 0.5779287 0.4302505 0.3601636
## [3,] 0.4095684 0.4640823 0.4176512 0.2413942
##
## , , 3, 2, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.08283823 0.2010452 0.2195096 0.08322205
## [2,] 0.20098587 0.3170443 0.2360299 0.19758115
## [3,] 0.22468397 0.2545896 0.2291181 0.13242579
##
## , , 1, 3, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.07486244 0.1816883 0.1983749 0.0752093
## [2,] 0.18163463 0.2865188 0.2133046 0.1785577
## [3,] 0.20305104 0.2300773 0.2070582 0.1196756
##
## , , 2, 3, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.04802152 0.1165464 0.1272502 0.04824403
## [2,] 0.11651199 0.1837913 0.1368271 0.11453827
## [3,] 0.13024984 0.1475862 0.1328203 0.07676755
##
## , , 3, 3, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2280494 0.5534671 0.6042985 0.2291060
## [2,] 0.5533037 0.8728065 0.6497782 0.5439307
## [3,] 0.6185433 0.7008719 0.6307502 0.3645613
##
## , , 1, 4, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1825495 0.4430407 0.4837303 0.1833953
## [2,] 0.4429099 0.6986662 0.5201360 0.4354069
## [3,] 0.4951330 0.5610356 0.5049044 0.2918249
##
## , , 2, 4, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1738329 0.4218859 0.4606327 0.1746384
## [2,] 0.4217613 0.6653056 0.4953000 0.4146167
## [3,] 0.4714909 0.5342467 0.4807957 0.2778906
##
## , , 3, 4, 4
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1184002 0.2873527 0.3137437 0.1189488
## [2,] 0.2872678 0.4531494 0.3373561 0.2824015
## [3,] 0.3211394 0.3638833 0.3274770 0.1892753
##
## , , 1, 1, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1558398 0.3782173 0.4129534 0.1565619
## [2,] 0.3781056 0.5964410 0.4440324 0.3717004
## [3,] 0.4226878 0.4789478 0.4310294 0.2491266
##
## , , 2, 1, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.03469573 0.08420523 0.09193878 0.03485649
## [2,] 0.08418036 0.13278994 0.09885812 0.08275434
## [3,] 0.09410601 0.10663159 0.09596318 0.05546485
##
## , , 3, 1, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2273704 0.5518193 0.6024993 0.2284239
## [2,] 0.5516563 0.8702078 0.6478435 0.5423112
## [3,] 0.6167017 0.6987852 0.6288722 0.3634759
##
## , , 1, 2, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1830860 0.4443429 0.4851521 0.1839344
## [2,] 0.4442117 0.7007197 0.5216648 0.4366867
## [3,] 0.4965883 0.5626846 0.5063884 0.2926827
##
## , , 2, 2, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.2093598 0.5081084 0.5547739 0.2103299
## [2,] 0.5079583 0.8012766 0.5965263 0.4993534
## [3,] 0.5678513 0.6434328 0.5790578 0.3346841
##
## , , 3, 2, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1730121 0.4198938 0.4584576 0.1738137
## [2,] 0.4197698 0.6621641 0.4929612 0.4126589
## [3,] 0.4692646 0.5317241 0.4785254 0.2765784
##
## , , 1, 3, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.008828473 0.02142637 0.02339420 0.008869379
## [2,] 0.021420040 0.03378895 0.02515486 0.021057183
## [3,] 0.023945662 0.02713285 0.02441823 0.014113260
##
## , , 2, 3, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1986243 0.4820537 0.5263263 0.1995446
## [2,] 0.4819113 0.7601889 0.5659378 0.4737477
## [3,] 0.5387332 0.6104390 0.5493650 0.3175223
##
## , , 3, 3, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.04992409 0.1211639 0.1322918 0.05015541
## [2,] 0.12112808 0.1910730 0.1422481 0.11907616
## [3,] 0.13541021 0.1534334 0.1380825 0.07980901
##
## , , 1, 4, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1474448 0.3578430 0.3907080 0.1481280
## [2,] 0.3577374 0.5643112 0.4201127 0.3516773
## [3,] 0.3999179 0.4531473 0.4078102 0.2357064
##
## , , 2, 4, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1407879 0.3416869 0.3730680 0.1414402
## [2,] 0.3415860 0.5388333 0.4011452 0.3357995
## [3,] 0.3818621 0.4326883 0.3893981 0.2250645
##
## , , 3, 4, 5
##
## [,1] [,2] [,3] [,4]
## [1,] 0.1628391 0.3952043 0.4315005 0.1635936
## [2,] 0.3950876 0.6232291 0.4639753 0.3883947
## [3,] 0.4416721 0.5004589 0.4503884 0.2603157DelayedTensor::einsum('ij,klm->ijklm', darrC, darrE)## <3 x 4 x 3 x 4 x 5> array of class HDF5Array and type "double":
## ,,1,1,1
## [,1] [,2] [,3] [,4]
## [1,] 0.09090300 0.22061813 0.24088009 0.09132419
## [2,] 0.22055298 0.34791033 0.25900877 0.21681679
## [3,] 0.24655823 0.27937532 0.25142401 0.14531819
##
## ,,2,1,1
## [,1] [,2] [,3] [,4]
## [1,] 0.03851239 0.09346811 0.10205239 0.03869084
## [2,] 0.09344051 0.14739733 0.10973287 0.09185762
## [3,] 0.10445802 0.11836146 0.10651948 0.06156619
##
## ,,3,1,1
## [,1] [,2] [,3] [,4]
## [1,] 0.06983767 0.16949338 0.18505995 0.07016126
## [2,] 0.16944333 0.26728764 0.19898760 0.16657294
## [3,] 0.18942227 0.21463452 0.19316049 0.11164301
##
## ...
##
## ,,1,4,5
## [,1] [,2] [,3] [,4]
## [1,] 0.1474448 0.3578430 0.3907080 0.1481280
## [2,] 0.3577374 0.5643112 0.4201127 0.3516773
## [3,] 0.3999179 0.4531473 0.4078102 0.2357064
##
## ,,2,4,5
## [,1] [,2] [,3] [,4]
## [1,] 0.1407879 0.3416869 0.3730680 0.1414402
## [2,] 0.3415860 0.5388333 0.4011452 0.3357995
## [3,] 0.3818621 0.4326883 0.3893981 0.2250645
##
## ,,3,4,5
## [,1] [,2] [,3] [,4]
## [1,] 0.1628391 0.3952043 0.4315005 0.1635936
## [2,] 0.3950876 0.6232291 0.4639753 0.3883947
## [3,] 0.4416721 0.5004589 0.4503884 0.2603157If there is a vanishing subscript on the left or right side of ->, the summation is done for that subscript.
einsum::einsum('i->', arrA)## [1] 0.8738674DelayedTensor::einsum('i->', darrA)## <1> array of class HDF5Array and type "double":
## [1]
## 0.8738674einsum::einsum('ij->', arrC)## [1] 6.624215DelayedTensor::einsum('ij->', darrC)## <1> array of class HDF5Array and type "double":
## [1]
## 6.624215einsum::einsum('ijk->', arrE)## [1] 32.28118DelayedTensor::einsum('ijk->', darrE)## <1> array of class HDF5Array and type "double":
## [1]
## 32.28118einsum::einsum('ij->i', arrC)## [1] 1.633352 2.649719 2.341144DelayedTensor::einsum('ij->i', darrC)## <3> array of class HDF5Array and type "double":
## [1] [2] [3]
## 1.633352 2.649719 2.341144einsum::einsum('ij->j', arrC)## [1] 1.415873 2.151422 1.906338 1.150581DelayedTensor::einsum('ij->j', darrC)## <4> array of class HDF5Array and type "double":
## [1] [2] [3] [4]
## 1.415873 2.151422 1.906338 1.150581einsum::einsum('ijk->i', arrE)## [1] 9.852426 11.231384 11.197368DelayedTensor::einsum('ijk->i', darrE)## <3> array of class HDF5Array and type "double":
## [1] [2] [3]
## 9.852426 11.231384 11.197368einsum::einsum('ijk->j', arrE)## [1] 6.252094 9.442822 7.062035 9.524228DelayedTensor::einsum('ijk->j', darrE)## <4> array of class HDF5Array and type "double":
## [1] [2] [3] [4]
## 6.252094 9.442822 7.062035 9.524228einsum::einsum('ijk->k', arrE)## [1] 6.504504 5.089863 7.233759 6.118140 7.334912DelayedTensor::einsum('ijk->k', darrE)## <5> array of class HDF5Array and type "double":
## [1] [2] [3] [4] [5]
## 6.504504 5.089863 7.233759 6.118140 7.334912These are the same as what the modeSum function does.
einsum::einsum('ijk->ij', arrE)## [,1] [,2] [,3] [,4]
## [1,] 2.773296 2.880774 1.579571 2.618785
## [2,] 1.304685 3.650797 2.544207 3.731695
## [3,] 2.174113 2.911251 2.938257 3.173748DelayedTensor::einsum('ijk->ij', darrE)## <3 x 4> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3] [,4]
## [1,] 2.773296 2.880774 1.579571 2.618785
## [2,] 1.304685 3.650797 2.544207 3.731695
## [3,] 2.174113 2.911251 2.938257 3.173748einsum::einsum('ijk->jk', arrE)## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.8638686 1.097706 1.321530 1.157144 1.811846
## [2,] 2.2173944 1.715838 1.676951 1.381076 2.451563
## [3,] 1.4613711 1.260435 1.702879 1.521484 1.115866
## [4,] 1.9618695 1.015885 2.532399 2.058436 1.955638DelayedTensor::einsum('ijk->jk', darrE)## <4 x 5> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.8638686 1.0977056 1.3215300 1.1571437 1.8118457
## [2,] 2.2173944 1.7158379 1.6769509 1.3810760 2.4515628
## [3,] 1.4613711 1.2604346 1.7028792 1.5214836 1.1158664
## [4,] 1.9618695 1.0158850 2.5323990 2.0584365 1.9556376einsum::einsum('ijk->jk', arrE)## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.8638686 1.097706 1.321530 1.157144 1.811846
## [2,] 2.2173944 1.715838 1.676951 1.381076 2.451563
## [3,] 1.4613711 1.260435 1.702879 1.521484 1.115866
## [4,] 1.9618695 1.015885 2.532399 2.058436 1.955638DelayedTensor::einsum('ijk->jk', darrE)## <4 x 5> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.8638686 1.0977056 1.3215300 1.1571437 1.8118457
## [2,] 2.2173944 1.7158379 1.6769509 1.3810760 2.4515628
## [3,] 1.4613711 1.2604346 1.7028792 1.5214836 1.1158664
## [4,] 1.9618695 1.0158850 2.5323990 2.0584365 1.9556376If we take the diagonal elements of a matrix
and add them together, we get trace.
einsum::einsum('ii->', arrB)## [1] 1.476876DelayedTensor::einsum('ii->', darrB)## <1> array of class HDF5Array and type "double":
## [1]
## 1.476876By changing the order of the indices on the left and right side of ->, we can get a sorted array or DelayedArray.
einsum::einsum('ij->ji', arrB)## [,1] [,2] [,3]
## [1,] 0.8486748 0.610428120 0.8016776
## [2,] 0.5055964 0.005582589 0.9735176
## [3,] 0.4763574 0.433875050 0.6226190DelayedTensor::einsum('ij->ji', darrB)## <3 x 3> matrix of class DelayedArray and type "double":
## [,1] [,2] [,3]
## [1,] 0.848674765 0.610428120 0.801677572
## [2,] 0.505596436 0.005582589 0.973517601
## [3,] 0.476357396 0.433875050 0.622618967einsum::einsum('ijk->jki', arrD)## , , 1
##
## [,1] [,2] [,3]
## [1,] 0.8355310 0.7260636 0.9468250
## [2,] 0.4983903 0.8234425 0.6118174
## [3,] 0.6909214 0.3490949 0.1348019
##
## , , 2
##
## [,1] [,2] [,3]
## [1,] 0.27195408 0.3570537 0.7993990
## [2,] 0.06854225 0.5134344 0.8875421
## [3,] 0.10257546 0.7728944 0.3233688
##
## , , 3
##
## [,1] [,2] [,3]
## [1,] 0.2368680 0.3383344 0.5961123
## [2,] 0.0444261 0.3617052 0.6819026
## [3,] 0.5979155 0.3627868 0.8110374DelayedTensor::einsum('ijk->jki', darrD)## <3 x 3 x 3> array of class DelayedArray and type "double":
## ,,1
## [,1] [,2] [,3]
## [1,] 0.8355310 0.7260636 0.9468250
## [2,] 0.4983903 0.8234425 0.6118174
## [3,] 0.6909214 0.3490949 0.1348019
##
## ,,2
## [,1] [,2] [,3]
## [1,] 0.27195408 0.35705370 0.79939901
## [2,] 0.06854225 0.51343445 0.88754213
## [3,] 0.10257546 0.77289439 0.32336877
##
## ,,3
## [,1] [,2] [,3]
## [1,] 0.2368680 0.3383344 0.5961123
## [2,] 0.0444261 0.3617052 0.6819026
## [3,] 0.5979155 0.3627868 0.8110374Some examples of combining Multiplication and Summation are shown below.
Inner Product first calculate Hadamard Product and collapses it to 0D tensor (norm).
einsum::einsum('i,i->', arrA, arrA)## [1] 0.2935518DelayedTensor::einsum('i,i->', darrA, darrA)## <1> array of class HDF5Array and type "double":
## [1]
## 0.2935518einsum::einsum('ij,ij->', arrC, arrC)## [1] 4.057632DelayedTensor::einsum('ij,ij->', darrC, darrC)## <1> array of class HDF5Array and type "double":
## [1]
## 4.057632einsum::einsum('ijk,ijk->', arrE, arrE)## [1] 21.74186DelayedTensor::einsum('ijk,ijk->', darrE, darrE)## <1> array of class HDF5Array and type "double":
## [1]
## 21.74186The inner product is an operation that eliminates all subscripts, while the outer product is an operation that leaves all subscripts intact. In the middle of the two, the operation that eliminates some subscripts while keeping others by summing them is called contracted product.
einsum::einsum('ijk,ijk->jk', arrE, arrE)## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.2748827 0.4681823 0.7455048 0.4723148 1.4508758
## [2,] 1.7284763 1.0718369 1.0169925 0.6924632 2.0166257
## [3,] 1.0742704 0.8558020 1.1131933 1.1262509 0.7898813
## [4,] 1.4329411 0.4977045 2.1761110 1.4578998 1.2796486DelayedTensor::einsum('ijk,ijk->jk', darrE, darrE)## <4 x 5> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.2748827 0.4681823 0.7455048 0.4723148 1.4508758
## [2,] 1.7284763 1.0718369 1.0169925 0.6924632 2.0166257
## [3,] 1.0742704 0.8558020 1.1131933 1.1262509 0.7898813
## [4,] 1.4329411 0.4977045 2.1761110 1.4578998 1.2796486Matrix Multiplication is considered a contracted product.
einsum::einsum('ij,jk->ik', arrC, t(arrC))## [,1] [,2] [,3]
## [1,] 0.7938124 1.152389 1.016463
## [2,] 1.1523885 1.827008 1.597971
## [3,] 1.0164628 1.597971 1.436811DelayedTensor::einsum('ij,jk->ik', darrC, t(darrC))## <3 x 3> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3]
## [1,] 0.7938124 1.1523885 1.0164628
## [2,] 1.1523885 1.8270079 1.5979714
## [3,] 1.0164628 1.5979714 1.4368113Some examples of combining Multiplication and Permutation are shown below.
einsum::einsum('ij,ij->ji', arrC, arrC)## [,1] [,2] [,3]
## [1,] 0.05320037 0.3131728 0.3913787
## [2,] 0.31335786 0.7792788 0.5024979
## [3,] 0.37355968 0.4319039 0.4069787
## [4,] 0.05369451 0.3026523 0.1359559DelayedTensor::einsum('ij,ij->ji', darrC, darrC)## <4 x 3> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3]
## [1,] 0.05320037 0.31317281 0.39137871
## [2,] 0.31335786 0.77927880 0.50249792
## [3,] 0.37355968 0.43190392 0.40697873
## [4,] 0.05369451 0.30265232 0.13595593einsum::einsum('ijk,ijk->jki', arrE, arrE)## , , 1
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.15532515 0.31772663 0.4525125 0.2182484 0.456501431
## [2,] 0.87265664 0.13005900 0.1805830 0.1348724 0.630080214
## [3,] 0.02121121 0.09979968 0.5702894 0.1053448 0.001465064
## [4,] 0.11905337 0.02319555 0.4771233 0.6263926 0.408643452
##
## , , 2
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.02787959 0.04174943 0.27827287 0.06525275 0.02262755
## [2,] 0.26439064 0.61776766 0.62135530 0.42860345 0.82389542
## [3,] 0.12368617 0.74982439 0.06617791 0.04334682 0.74156668
## [4,] 0.56252392 0.43191583 0.92241825 0.56800153 0.37257694
##
## , , 3
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.09167795 0.108706276 0.01471947 0.1888136 0.97174682
## [2,] 0.59142900 0.324010259 0.21505423 0.1289873 0.56265005
## [3,] 0.92937299 0.006177919 0.47672598 0.9775592 0.04684957
## [4,] 0.75136385 0.042593170 0.77656948 0.2635057 0.49842825DelayedTensor::einsum('ijk,ijk->jki', darrE, darrE)## <4 x 5 x 3> array of class HDF5Array and type "double":
## ,,1
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.155325155 0.317726628 0.452512462 0.218248413 0.456501431
## [2,] 0.872656637 0.130059001 0.180582969 0.134872425 0.630080214
## [3,] 0.021211206 0.099799681 0.570289399 0.105344836 0.001465064
## [4,] 0.119053370 0.023195547 0.477123251 0.626392608 0.408643452
##
## ,,2
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.02787959 0.04174943 0.27827287 0.06525275 0.02262755
## [2,] 0.26439064 0.61776766 0.62135530 0.42860345 0.82389542
## [3,] 0.12368617 0.74982439 0.06617791 0.04334682 0.74156668
## [4,] 0.56252392 0.43191583 0.92241825 0.56800153 0.37257694
##
## ,,3
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.091677947 0.108706276 0.014719470 0.188813636 0.971746820
## [2,] 0.591429003 0.324010259 0.215054231 0.128987309 0.562650046
## [3,] 0.929372992 0.006177919 0.476725978 0.977559223 0.046849572
## [4,] 0.751363852 0.042593170 0.776569477 0.263505657 0.498428248Some examples of combining Summation and Permutation are shown below.
einsum::einsum('ijk->ki', arrE)## [,1] [,2] [,3]
## [1,] 1.818956 1.782868 2.902680
## [2,] 1.392521 2.513436 1.183906
## [3,] 2.543557 2.533453 2.156748
## [4,] 1.950439 1.871982 2.295719
## [5,] 2.146953 2.529645 2.658314DelayedTensor::einsum('ijk->ki', darrE)## <5 x 3> matrix of class HDF5Matrix and type "double":
## [,1] [,2] [,3]
## [1,] 1.818956 1.782868 2.902680
## [2,] 1.392521 2.513436 1.183906
## [3,] 2.543557 2.533453 2.156748
## [4,] 1.950439 1.871982 2.295719
## [5,] 2.146953 2.529645 2.658314Finally, we will show a more complex example, combining Multiplication, Summation, and Permutation.
einsum::einsum('i,ij,ijk,ijk,ji->jki',
arrA, arrC, arrE, arrE, t(arrC))## , , 1
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.003664946 0.0074968599 0.01067717 0.00514964 0.0107712951
## [2,] 0.121281657 0.0180755757 0.02509739 0.01874454 0.0875684308
## [3,] 0.003514281 0.0165348510 0.09448578 0.01745357 0.0002427323
## [4,] 0.002835193 0.0005523898 0.01136244 0.01491721 0.0097316291
##
## , , 2
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.002280889 0.003415611 0.022766105 0.005338469 0.001851209
## [2,] 0.053823601 0.125762698 0.126493055 0.087253396 0.167725372
## [3,] 0.013955394 0.084601978 0.007466791 0.004890780 0.083670268
## [4,] 0.044475294 0.034148918 0.072929916 0.044908374 0.029457358
##
## , , 3
##
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.006067909 0.0071949673 0.0009742409 0.01249705 0.06431723
## [2,] 0.050259006 0.0275340463 0.0182750793 0.01096120 0.04781340
## [3,] 0.063964464 0.0004251978 0.0328108542 0.06728090 0.00322444
## [4,] 0.017275295 0.0009792986 0.0178548205 0.00605850 0.01145982DelayedTensor::einsum('i,ij,ijk,ijk,ji->jki',
darrA, darrC, darrE, darrE, t(darrC))## <4 x 5 x 3> array of class HDF5Array and type "double":
## ,,1
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.0036649460 0.0074968599 0.0106771741 0.0051496401 0.0107712951
## [2,] 0.1212816569 0.0180755757 0.0250973875 0.0187445445 0.0875684308
## [3,] 0.0035142811 0.0165348510 0.0944857755 0.0174535746 0.0002427323
## [4,] 0.0028351934 0.0005523898 0.0113624395 0.0149172108 0.0097316291
##
## ,,2
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.002280889 0.003415611 0.022766105 0.005338469 0.001851209
## [2,] 0.053823601 0.125762698 0.126493055 0.087253396 0.167725372
## [3,] 0.013955394 0.084601978 0.007466791 0.004890780 0.083670268
## [4,] 0.044475294 0.034148918 0.072929916 0.044908374 0.029457358
##
## ,,3
## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.0060679094 0.0071949673 0.0009742409 0.0124970515 0.0643172303
## [2,] 0.0502590061 0.0275340463 0.0182750793 0.0109612040 0.0478134010
## [3,] 0.0639644642 0.0004251978 0.0328108542 0.0672809006 0.0032244404
## [4,] 0.0172752949 0.0009792986 0.0178548205 0.0060585000 0.0114598206einsumBy using einsum and other DelayedTensor functions,
it is possible to implement your original tensor calculation functions.
It is intended to be applied to Delayed Arrays,
which can scale to large-scale data
since the calculation is performed internally by block processing.
For example, kronecker can be easily implmented by eimsum
and other DelayedTensor functions4 https://stackoverflow.com/
questions/56067643/speeding-up-kronecker-products-numpy
(the kronecker function inside DelayedTensor
has a more efficient implementation though).
darr1 <- DelayedArray(array(1:6, dim=c(2,3)))
darr2 <- DelayedArray(array(20:1, dim=c(4,5)))
mykronecker <- function(darr1, darr2){
stopifnot((length(dim(darr1)) == 2) && (length(dim(darr2)) == 2))
# Outer Product
tmpdarr <- DelayedTensor::einsum('ij,kl->ikjl', darr1, darr2)
# Reshape
DelayedTensor::unfold(tmpdarr, row_idx=c(2,1), col_idx=c(4,3))
}
identical(as.array(DelayedTensor::kronecker(darr1, darr2)),
as.array(mykronecker(darr1, darr2)))## [1] TRUE## R version 4.3.0 RC (2023-04-13 r84269)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 22.04.2 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.17-bioc/R/lib/libRblas.so
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.10.0
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_GB LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## time zone: America/New_York
## tzcode source: system (glibc)
##
## attached base packages:
## [1] stats4 stats graphics grDevices utils datasets methods
## [8] base
##
## other attached packages:
## [1] einsum_0.1.0 DelayedRandomArray_1.8.0 HDF5Array_1.28.0
## [4] rhdf5_2.44.0 DelayedArray_0.26.0 IRanges_2.34.0
## [7] S4Vectors_0.38.0 MatrixGenerics_1.12.0 matrixStats_0.63.0
## [10] BiocGenerics_0.46.0 Matrix_1.5-4 DelayedTensor_1.6.0
## [13] BiocStyle_2.28.0
##
## loaded via a namespace (and not attached):
## [1] jsonlite_1.8.4 compiler_4.3.0 BiocManager_1.30.20
## [4] Rcpp_1.0.10 rsvd_1.0.5 rhdf5filters_1.12.0
## [7] parallel_4.3.0 jquerylib_0.1.4 BiocParallel_1.34.0
## [10] yaml_2.3.7 fastmap_1.1.1 lattice_0.21-8
## [13] R6_2.5.1 ScaledMatrix_1.8.0 knitr_1.42
## [16] bookdown_0.33 bslib_0.4.2 rlang_1.1.0
## [19] cachem_1.0.7 xfun_0.39 sass_0.4.5
## [22] cli_3.6.1 Rhdf5lib_1.22.0 BiocSingular_1.16.0
## [25] digest_0.6.31 grid_4.3.0 irlba_2.3.5.1
## [28] rTensor_1.4.8 dqrng_0.3.0 evaluate_0.20
## [31] codetools_0.2-19 beachmat_2.16.0 rmarkdown_2.21
## [34] tools_4.3.0 htmltools_0.5.5