Merge branch 'dev' of https://github.com/virtual-labs/exp-mean-field-…

…annealing-iiith into testing

experiment-descriptor.json

@@ -0,0 +1,55 @@
+{
+  "unit-type": "lu",
+  "label": "",
+  "basedir": ".",
+  "units": [
+    {
+      "unit-type": "aim"
+    },
+    {
+      "target": "theory.html",
+      "source": "theory.md",
+      "label": "Theory",
+      "unit-type": "task",
+      "content-type": "text"
+    },
+
+    {
+      "target": "procedure.html",
+      "source": "procedure.md",
+      "label": "Procedure",
+      "unit-type": "task",
+      "content-type": "text"
+    },
+    {
+      "target": "simulation.html",
+      "source": "simulation/index.html",
+      "label": "Simulation",
+      "unit-type": "task",
+      "content-type": "simulation"
+    },
+	{
+      "target": "observations.html",
+      "source": "observations.md",
+      "label": "Observations",
+      "unit-type": "task",
+      "content-type": "text"
+    },
+    {
+      "target": "assignment.html",
+      "source": "assignment.md",
+      "label": "Assignment",
+      "unit-type": "task",
+      "content-type": "text"
+    },
+
+    {
+      "target": "references.html",
+      "source": "references.md",
+      "label": "References",
+      "unit-type": "task",
+      "content-type": "text"
+    }
+  ]
+}
+

experiment/experiment-name.md

@@ -1 +1 @@
-## Deterministic, stochastic and mean-field annealing of Hopfield models
+## Deterministic, Stochastic and Mean-field Annealing of Hopfield Models

experiment/simulation/backend/Dockerfile

@@ -0,0 +1,12 @@
+FROM ubuntu:18.04
+
+RUN apt update && \
+    apt install -y octave python3 python3-pip
+
+RUN octave --eval 'pkg install -forge nnet'
+
+COPY . /
+
+RUN pip3 install -r requirements.txt
+
+CMD python3 server.py

experiment/simulation/backend/exp_weighted_matching/__init__.py

@@ -0,0 +1,48 @@
+from utils import *
+
+EXP_NAME = "weighted_matching"
+
+
+def init_graph(nodeloc, token):
+    clear_files(10, token)
+
+    res = run_octave(EXP_NAME, "init_wmp", nodeloc, token)
+
+    if res:
+        return res
+
+    images = read_all_pngs(EXP_NAME, token, 1)
+
+    return jsonify(images=images, result="")
+
+
+def start_annealing(method, nodeloc, token):
+    res = run_octave(EXP_NAME, f"anneal_wmp_{method}", nodeloc, token)
+
+    if res:
+        return res
+
+    images = read_all_pngs(EXP_NAME, token)
+    result = read_result(EXP_NAME, token)
+
+    return jsonify(images=images, result=result)
+
+
+num_map = {
+    "det": "det",
+    "mean": "mean",
+    "sto": "sto",
+}
+
+
+def runexp(data, part_num):
+    nodeloc = data.get("nodeloc")
+    token = data.get("token")
+
+    if data.get("annealflag") == "1":
+        return start_annealing(num_map[part_num], nodeloc, token)
+
+    if data.get("initflag") == "1":
+        return init_graph(nodeloc, token)
+
+    return jsonify(success=True)

experiment/simulation/backend/exp_weighted_matching/anneal_wmp_det.m

@@ -0,0 +1,95 @@
+function finalstate = anneal_wmp_det(points, filename, initstate)
+    %%%%%%%%%
+    % Ex1: points=[1 1;3 3;0 4;2 6];
+    %   The network yields solution only for certain initial states. For other
+    %   init states, the network oscillates or gets stuck at some local minima.
+    % validinitstates=[5 6 7 9 10 11 13 14 15 21 25 29 37 41 45];
+    %%%%%%%%%
+
+    fid=fopen(strcat(filename, "-result.txt"),'w');
+
+    N=length(points);
+    M=nchoosek(N,2);
+    if(~exist('initstate'))
+        initstate=(rand(M,1)-0.5)>0;
+    end
+    initstate=initstate(:);
+
+    fprintf(fid,'Initial state:\n');
+    fprintf(fid,'%d ',initstate);
+    fprintf(fid,'\n');
+
+    d=zeros(N,N);
+    unit=[];
+    th=[];
+    k=1;
+    for i=1:N
+        for j=i+1:N
+            d(i,j)=norm(points(i,:)-points(j,:));
+            unit(k,:)=[i j];
+            th=[th; d(i,j)];
+            k=k+1;
+        end
+    end
+    d=d+d';
+
+
+    % The nodes are (i,j): 12, 13, 14, 23, 24, 34
+
+    gamma=max(max(d))/(M/2);
+    W = zeros(M,M);
+
+    for i=1:M
+        for j=i+1:M
+            if(any(~[unit(i,:)-unit(j,:) unit(i,[2 1])-unit(j,:)]))
+                W(i,j)=-gamma;
+            end
+        end
+    end
+
+    W = W + W';
+
+    th=th-gamma;
+
+    % Random initialization
+    %state = (rand(M,1)-0.5)>0;
+    %state=[0 1 1 0 0 0]'
+    state=initstate;
+    disp(state)
+    tempstate = zeros(M,1);
+    stateseq=[zeros(M,2) state];
+    k=3;
+    while(norm(tempstate-state) ~= 0 & norm(stateseq(:,k)-stateseq(:,k-2))~=0 | k<10)
+        tempstate=state;
+        state=W*state+th;
+        %disp(state')
+        state=state>0;
+        %disp(state')
+        stateseq=[stateseq state];
+        k=k+1;
+        %pause;
+    end
+    finalstate=state;
+
+    disp(stateseq);
+    fprintf(fid,'\nOutput state of the nodes after each update:\n');
+    for i=3:size(stateseq,2)
+        fprintf(fid,'%d ',stateseq(:,i));
+        fprintf(fid,'\n');
+    end
+
+
+    g=zeros(N,N);
+    ndx=unit(find(finalstate),:);
+    for i=1:size(ndx,1)
+        g([ndx(i,1)],[ndx(i,2)])=1;
+    end
+    %xyminmax = min_max(points');
+
+    drawgraph(g,points);
+
+    print(strcat(filename, "-2.png"))
+
+    fclose(fid);
+
+    return;

experiment/simulation/backend/exp_weighted_matching/anneal_wmp_mean.m

@@ -0,0 +1,113 @@
+function FinalState = anneal_wmp_mean(points, filename, InitState)
+    %%%%%%%%%
+    % Ex1: points=[1 1;3 3;0 4;2 6];
+    %   The network yields solution only for certain initial states. For other
+    %   init states, the network oscillates or gets stuck at some local minima.
+    % validinitstates=[5 6 7 9 10 11 13 14 15 21 25 29 37 41 45];
+    %%%%%%%%%
+
+    fid=fopen(strcat(filename, "-result.txt"),'w');
+
+    %points=[1 1;3 3;0 4;2 6];
+    N=length(points);
+    M=nchoosek(N,2);
+
+    if(~exist('InitState'))
+        InitState=rand(M,1);
+    end
+
+    fprintf(fid,'Initial state:\n');
+    fprintf(fid,'%d ',InitState);
+    fprintf(fid,'\n');
+    fprintf(fid,'\n');
+    fprintf(fid,'\nAvg. output state of the nodes after annealing for each temperature:\n');
+
+    d=zeros(N,N);
+    unit=[];
+    th=[];
+    k=1;
+    for i=1:N
+        for j=i+1:N
+            d(i,j)=norm(points(i,:)-points(j,:));
+            unit(k,:)=[i j];
+            th=[th; d(i,j)];
+            k=k+1;
+        end
+    end
+    d=d+d';
+
+
+    % The nodes are (i,j): 12, 13, 14, 23, 24, 34
+
+    gamma=max(max(d))/(M/2);
+    W = zeros(M,M);
+
+    for i=1:M
+        for j=i+1:M
+            if(any(~[unit(i,:)-unit(j,:) unit(i,[2 1])-unit(j,:)]))
+                W(i,j)=-gamma;
+            end
+        end
+    end
+
+    W = W + W';
+
+    th=th-gamma;
+    th=th(:);
+
+    % Random initialization
+    %State = (rand(M,1)-0.5)>0;
+    %state=[0 1 1 0 0 0]'
+
+    State=InitState(:);
+    TempState=zeros(M,1);
+    disp(State')
+
+    T=1;
+    dT=0.01;
+    i=1;
+    while(T>0)
+        %TempState=rand(M,1);
+        T=T-dT;
+        while(norm(State-TempState) >= 0.00001 & i<100)
+            TempState = State;
+            State = (W*State + th);
+            State = tanh(State/T);
+            i=i+1;
+            %disp(State')
+            %disp(norm(State-TempState))
+            %pause;
+        end
+        disp(State')
+
+        fprintf(fid,'%4.2f :',T);
+        fprintf(fid,'%d ',State>0);
+        fprintf(fid,'\n');
+
+    end
+
+    [sval,spos]=sort(State,'descend');
+    FinalState=zeros(M,1);
+    FinalState(spos(1:floor(N/2)))=1;
+
+    disp(FinalState)
+    fprintf(fid,'\n');
+    fprintf(fid,'\nFinal output state of the nodes:\n');
+    fprintf(fid,'%d ',FinalState);
+    fprintf(fid,'\n');
+
+
+    g=zeros(N,N);
+    ndx=unit(find(FinalState),:);
+    for i=1:size(ndx,1)
+        g([ndx(i,1)],[ndx(i,2)])=1;
+    end
+    %xyminmax = min_max(points');
+
+    drawgraph(g,points);
+
+    print(strcat(filename, "-2.png"))
+
+    fclose(fid);
+
+    return;