Skip to content
Snippets Groups Projects
Commit 1b579074 authored by Anuarbekov, Alikhan's avatar Anuarbekov, Alikhan
Browse files

Final version, debug/comments removed

parent 268a61f7
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
data/hsa_circ_0000228_BN.RData
+ 0
− 0
View file @ 1b579074
No preview for this file type
source/skeleton.cpp
+ 4
− 112
View file @ 1b579074
......@@ -121,12 +121,6 @@ enum ChangeType
struct GraphChange
{
/**
* Reference to parent for potential traversal
*/
//GraphChange* prev = nullptr;
/**
* Note: since unordered_map does not have any guarantee about the key order,
* the random sampling is modified by the changing key ordering!
......@@ -141,13 +135,6 @@ struct GraphChange
>
next_map;
/**
* key_tp, but divided to be more readable
*/
//uint32_t from;
//uint32_t to;
//uint32_t listID;
/**
* After Trie receives a word, only the end of it is actually a subgraph,
*
......@@ -181,64 +168,12 @@ struct GraphChange
{
this->is_graph = true;
}
/*
key_tp key()
{
return std::make_tuple(this->listID, this->from, this->to);
}*/
//key_tp kt, GraphChange* parent
GraphChange()
{
//this->listID = std::get<0>(kt);
//this->from = std::get<1>(kt);
//this->to = std::get<2>(kt);
this->is_graph = false;
//this->prev = parent;
}
/**
* When going down the node, we add the edge, storing 1 ADJ matrix only
*/
/*
GraphChange* forward(std::vector<LogicalMatrix*>& adjs)
{
(*adjs[listID])(from, to) = true;
return next;
}*/
/**
* When going up the node, we remove the edge, storing 1 ADJ matrix only
*/
/*
GraphChange* backward(std::vector<LogicalMatrix*>& adjs)
{
(*adjs[listID])(from, to) = false;
return prev;
}
void printG()
{
Rcout << listID << "," << from << "," << to << "; " << is_graph << "\n";
}*/
};
/*
struct graph_comp
{
inline bool operator() (GraphChange*& k1, GraphChange*& k2)
{
key_tp t1 = std::make_tuple(k1->listID, k1->from, k1->to);
key_tp t2 = std::make_tuple(k2->listID, k2->from, k2->to);
return key_comp()(t1, t2);
}
};*/
class Trie
{
......@@ -331,7 +266,7 @@ public:
this->find(&currIndex, sampledIndex, G_adjs);
}
//
/**
* Returns a number of subgraphs in Trie.
*
......@@ -377,15 +312,13 @@ protected:
// Note: if the sampled subgraphs are too deep and stack memory overflow,
// then consider using the non-recursive version
//for(const auto & [ key, value ] : this->root)
//{
for(const auto &item : this->root)
{
const auto& key = item.first;
const auto& value = item.second;
// Go down
//value->forward(G_adjs);
(*G_adjs[std::get<0>(key)])(std::get<1>(key), std::get<2>(key)) = true;
*currIdx += value->is_graph;
......@@ -397,7 +330,6 @@ protected:
break;
}
//value->backward(G_adjs);
(*G_adjs[std::get<0>(key)])(std::get<1>(key), std::get<2>(key)) = false;
}
}
......@@ -410,11 +342,6 @@ protected:
return;
}
//curr->forward(G_adjs);
//currIdx += curr->is_graph;
//for(const auto & [ key, value ] : curr->next_map)
//{
for(const auto &item : this->root)
{
const auto& key = item.first;
......@@ -430,8 +357,6 @@ protected:
(*G_adjs[std::get<0>(key)])(std::get<1>(key), std::get<2>(key)) = false;
}
//curr->backward(G_adjs);
return;
}
......@@ -548,8 +473,7 @@ public:
List& result()
{
Rprintf("Optimum log probability: %.10f\n", this->probabilityMaxOptim);
//Rprintf("Trie total size: %.5d\n", this->V_G.total_num());
const List* SK_list = this->input_data.SK;
List* ret = new List(SK_list->length());
......@@ -577,7 +501,6 @@ protected:
{
Rprintf("%5d/%5d \n Array:20%d\n", i, SK_list->length(), this->unord_V_G.size());
uint32_t l = ls[i] - 1;
//if(l == 0) continue;
const IntegerMatrix& sk = (*SK_list)[l];
// Random permutation of the indices
......@@ -603,12 +526,6 @@ protected:
uint32_t r = lin % sk.nrow();
uint32_t c = lin / sk.nrow();
// Rprintf("\014[pSK-SS][add only][%5u,%5u,%5u]/[%5u,%5u,%5u]\n"
// "Probability optim: %.10f\nTrie: nodes: %.10lu\n, subgraphs %.5u\n",
// i,j,k,
// SK_list->length(), (uint32_t)sk.ncol(), (uint32_t)sk.nrow(),
// this->probabilityMaxOptim, this->V_G.nodes_num(), this->V_G.total_num());
// if inside G_0 -> skip
if((*this->G[l])(r,c) != 0)
continue;
......@@ -640,13 +557,10 @@ protected:
const IntegerMatrix& sk = (*SK_list)[l];
Rprintf("%5d/%5d \n Array:20%d\n", i, SK_list->length(), this->unord_V_G.size());
//if(l == 0) continue;
// Random permutation of the indices
uint32_t num = sk.ncol() * sk.nrow();
IntegerVector xy = Rcpp::sample(num, num);
//IntegerVector xy = Rcpp::Range(1, num); // TODO remove
// It is necessary to do 2d loop in COL, ROW order, since
// rows are local in memory by Rcpp standard
// UPD: after random permutation this doesn't matter
......@@ -661,13 +575,6 @@ protected:
uint32_t r = lin % sk.nrow();
uint32_t c = lin / sk.nrow();
// Rprintf("\014[pSK-SS][add only][%5u,%5u,%5u]/[%5u,%5u,%5u]\n"
// "Probability optim: %.10f\nTrie: nodes: %.10lu\n, subgraphs %.5u\n",
// i,j,k,
// SK_list->length(), (uint32_t)sk.ncol(), (uint32_t)sk.nrow(),
// this->probabilityMaxOptim, this->V_G.nodes_num(), this->V_G.total_num()
// );
if(sk(r, c) == 0)
{
continue;
......@@ -676,7 +583,6 @@ protected:
this->tryPerformMove(&twoScores, l, r, c, ChangeType::ADD);
}
}
Rcout << "GLOBAL COUNTER: " << this->ord_V_G.size() << "\n";
}
Rprintf("partialSKSS-add-remove phase\n");
......@@ -722,21 +628,12 @@ protected:
continue;
}
// Rprintf("\014[pSK-SS][add only][%5u,%5u,%5u]/[%5u,%5u,%5u]\n"
// "Probability optim: %.10f\nTrie: nodes: %.10lu\n, subgraphs %.5u\n",
// i,j,k,
// SK_list->length(), (uint32_t)gz.ncol(), (uint32_t)gz.nrow(),
// this->probabilityMaxOptim, this->V_G.nodes_num(), this->V_G.total_num());
if(gz(r,c) == 0) // if not in G -> try add
this->tryPerformMove(&twoScores, l, r, c, ChangeType::ADD);
else // if in G -> try delete
this->tryPerformMove(&twoScores, l, r, c, ChangeType::DELETE);
}
}
}
}
......@@ -999,8 +896,6 @@ List c_SK_SS(
Function score_change_func
)
{
Rcout << "I am in SK_SS function!\n";
IData i_data;
i_data.PK = &PK;
......@@ -1022,9 +917,6 @@ List c_SK_SS(
net->SK_SS();
List ret = net->result();
delete net;
return ret;
......
workflow.Rmd
+ 42
− 9
View file @ 1b579074
......@@ -412,6 +412,8 @@ MDS_RELATED_TERMS <- convert_circGPA(res[[2]])
```{r}
library(ggVennDiagram)
library(reshape2)
library(ggplot2)
rank1_circGPA <- as.character(df1_orig$Row.names[order(df1_orig$pvalue)])
rank1_circGPA <- rank1_circGPA[1:100]
......@@ -445,9 +447,22 @@ rank1_BN <- as.character(df1$Row.names[order(df1$pvalue)])
rank1_BN_ids <- match(MDS_RELATED_TERMS, rank1_BN)
rank1_BN_ids <- rank1_BN_ids[!is.na(rank1_BN_ids)]
bdata <- data.frame(orig=rank1_circGPA_ids, corr=rank1_corr_ids, BN=rank1_BN_ids)
boxplot(bdata$orig, bdata$corr, bdata$BN,
names = c("circGPA", "circGPAcorr", "circGPA-BN"))
# bdata <- data.frame(orig=rank1_circGPA_ids, corr=rank1_corr_ids, BN=rank1_BN_ids)
# boxplot(bdata$orig, bdata$corr, bdata$BN,
# names = c("circGPA", "circGPAcorr", "circGPA-BN"))
bdata <- data.frame(
type=c(
rep("circGPA",length(rank1_circGPA_ids)),
rep("circGPAcorr",length(rank1_corr_ids)),
rep("circGPA-BN",length(rank1_BN_ids))
),
data=c(
rank1_circGPA_ids, rank1_corr_ids, rank1_BN_ids
)
)
ggplot(data = bdata, aes(x=data, y=type)) + geom_boxplot(aes(fill=type)) + xlab("Position in GO term ordering") + ylab("circGPA version") + ggtitle("A distribution of MDS-related GO terms")
```
......@@ -482,9 +497,18 @@ rank2_BN <- as.character(df2$Row.names[order(df2$pvalue)])
rank2_BN_ids <- match(MDS_RELATED_TERMS, rank2_BN)
rank2_BN_ids <- rank2_BN_ids[!is.na(rank2_BN_ids)]
bdata <- data.frame(orig=rank2_circGPA_ids, corr=rank2_corr_ids, BN=rank2_BN_ids)
boxplot(bdata$orig, bdata$corr, bdata$BN,
names = c("circGPA", "circGPAcorr", "circGPA-BN"))
bdata <- data.frame(
type=c(
rep("circGPA",length(rank2_circGPA_ids)),
rep("circGPAcorr",length(rank2_corr_ids)),
rep("circGPA-BN",length(rank2_BN_ids))
),
data=c(
rank2_circGPA_ids, rank2_corr_ids, rank2_BN_ids
)
)
ggplot(data = bdata, aes(x=data, y=type)) + geom_boxplot(aes(fill=type)) + xlab("Position in GO term ordering") + ylab("circGPA version") + ggtitle("A distribution of MDS-related GO terms")
```
```{r}
......@@ -518,9 +542,18 @@ rank3_BN <- as.character(df3$Row.names[order(df3$pvalue)])
rank3_BN_ids <- match(MDS_RELATED_TERMS, rank3_BN)
rank3_BN_ids <- rank3_BN_ids[!is.na(rank3_BN_ids)]
bdata <- data.frame(orig=rank3_circGPA_ids, corr=rank3_corr_ids, BN=rank3_BN_ids)
boxplot(bdata$orig, bdata$corr, bdata$BN,
names = c("circGPA", "circGPAcorr", "circGPA-BN"))
bdata <- data.frame(
type=c(
rep("circGPA",length(rank3_circGPA_ids)),
rep("circGPAcorr",length(rank3_corr_ids)),
rep("circGPA-BN",length(rank3_BN_ids))
),
data=c(
rank3_circGPA_ids, rank3_corr_ids, rank3_BN_ids
)
)
ggplot(data = bdata, aes(x=data, y=type)) + geom_boxplot(aes(fill=type)) + xlab("Position in GO term ordering") + ylab("circGPA version") + ggtitle("A distribution of MDS-related GO terms")
```
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment