fix typo

Project.toml

@@ -1,7 +1,7 @@
 name = "Stuffing"
 uuid = "4175e07e-e5b7-423e-8796-3ea7f6d48281"
 authors = ["guoyongzhi <[email protected]>"]
-version = "0.8.4"
+version = "0.8.5"
 
 [deps]
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"

README.md

@@ -51,5 +51,5 @@ import Pkg; Pkg.add("Stuffing")
 ![pyramid2](./res/pyramid2.png)
 * Second, a **top-down method** ([`collision`](./src/qtree_functions.jl)) is used to detect collision between two pyramids (trees). On the level 𝑙 and coordinates (𝑎,𝑏), if one tree's node is `FULL` and the other's is NOT `EMPTY`, then these two trees collide at (𝑙,𝑎,𝑏). However, to detect collisions between many objects, pairwise detection would be time-consuming. So, we first locate the objects in hierarchical sub-regions (implemented as a [`HashSpacialQTree` or `LinkedSpacialQTree`](./src/qtree.jl)), and then detect the collision between objects within each sub-region and between the objects in the sub-regions and those in their ancestral regions (see [`totalcollisions_spacial`, `partialcollisions` and `locate!`](./src/qtree_functions.jl)).  
 ![collision](./res/collision.png)
-* At last, each object in collision pair is moved according to the **local gradient** (calculated by [`gard2d`](./src/fit.jl)) near the collision point (𝑙,𝑎,𝑏), that is, moving the object away from `EMPTY` region. This will enlarge space between them. After moving the objects, the `AbstractStackedQTree`s should be rebuilt for the next round of collision detection.  
+* At last, each object in collision pair is moved according to the **local gradient** (calculated by [`grad2d`](./src/fit.jl)) near the collision point (𝑙,𝑎,𝑏), that is, moving the object away from `EMPTY` region. This will enlarge space between them. After moving the objects, the `AbstractStackedQTree`s should be rebuilt for the next round of collision detection.  
 ![gradient](./res/gradient.png)

src/fit.jl

@@ -29,8 +29,8 @@ function step!(t1, t2, collisionpoint::Tuple{Integer,Integer,Integer}, optimiser
     ks2 = (ks2[1] + ks2[2]) ^ 3
     l = collisionpoint[1]
     ll = 2^(l - 1)
-    delta1 = ll .* gard2d(t1, collisionpoint...)
-    delta2 = ll .* gard2d(t2, collisionpoint...)
+    delta1 = ll .* grad2d(t1, collisionpoint...)
+    delta2 = ll .* grad2d(t2, collisionpoint...)
     move1 = rand() < ks2 / ks1 # ks1越大移动概率越小，ks1<=ks2时必然移动（质量越大，惯性越大运动越少）
     move2 = rand() < ks1 / ks2
     if move1
@@ -49,8 +49,8 @@ end
 function step_mask!(mask, t2, collisionpoint::Tuple{Integer,Integer,Integer}, optimiser=(t, Δ) -> Δ ./ 6)
     l = collisionpoint[1]
     ll = 2^(l - 1)
-    delta1 = ll .* gard2d(mask, collisionpoint...)
-    delta2 = ll .* gard2d(t2, collisionpoint...)
+    delta1 = ll .* grad2d(mask, collisionpoint...)
+    delta2 = ll .* grad2d(t2, collisionpoint...)
     delta2 = (delta2 .- delta1) ./ 2
     move!(t2, optimiser(t2, delta2))
 end

src/fit_helper.jl

@@ -45,10 +45,10 @@ eta!(o::SGD, e) = o.eta = e
 # near(a::Integer, b::Integer, r=1) = a-r:a+r, b-r:b+r
 # near(m::AbstractMatrix, a::Integer, b::Integer, r=1) = @view m[near(a, b, r)...]
 # const KERNEL = collect.(Iterators.product(-1:1,-1:1))
-# gard2d(m::AbstractMatrix) = sum(KERNEL .* m)
-# gard2d(t::ShiftedQTree, l, a, b) = gard2d(decode2(near(t[l],a,b)))|>Tuple
+# grad2d(m::AbstractMatrix) = sum(KERNEL .* m)
+# grad2d(t::ShiftedQTree, l, a, b) = grad2d(decode2(near(t[l],a,b)))|>Tuple
 
-function gard2d(t::ShiftedQTree, l, a, b) # FULL is white, Positive directions are right & down 
+function grad2d(t::ShiftedQTree, l, a, b) # FULL is white, Positive directions are right & down 
     m = t[l]
     diag = -decode2(m[a - 1, b - 1]) + decode2(m[a + 1, b + 1])
     cdiag = -decode2(m[a - 1, b + 1]) + decode2(m[a + 1, b - 1])