358 lines
13 KiB
Python
358 lines
13 KiB
Python
import networkx as nx
|
|
from scipy.stats import expon, gamma
|
|
import numpy as np
|
|
import matplotlib.pyplot as plt
|
|
import matplotlib.colors as colors
|
|
import matplotlib.cm as cmx
|
|
|
|
#helper functions
|
|
def get_nodes_by_type(g, node_type_selection):
|
|
return [node for node in g.nodes if g.nodes[node]['type']== node_type_selection ]
|
|
|
|
def get_edges_by_type(g, edge_type_selection):
|
|
return [edge for edge in g.edges if g.edges[edge]['type']== edge_type_selection ]
|
|
|
|
def total_funds_given_total_supply(total_supply):
|
|
|
|
#can put any bonding curve invariant here for initializatio!
|
|
total_funds = total_supply**2/1000
|
|
|
|
return total_funds
|
|
|
|
#maximum share of funds a proposal can take
|
|
default_beta = .2 #later we should set this to be param so we can sweep it
|
|
# tuning param for the trigger function
|
|
default_rho = .001
|
|
|
|
def trigger_threshold(requested, funds, supply, beta = default_beta, rho = default_rho):
|
|
|
|
share = requested/funds
|
|
if share < beta:
|
|
return rho*supply/(beta-share)**2
|
|
else:
|
|
return np.inf
|
|
|
|
def initialize_network(n,m, funds_func=total_funds_given_total_supply, trigger_func =trigger_threshold ):
|
|
network = nx.DiGraph()
|
|
for i in range(n):
|
|
network.add_node(i)
|
|
network.nodes[i]['type']="participant"
|
|
|
|
h_rv = expon.rvs(loc=0.0, scale=1000)
|
|
network.nodes[i]['holdings'] = h_rv
|
|
|
|
s_rv = np.random.rand()
|
|
network.nodes[i]['sentiment'] = s_rv
|
|
|
|
participants = get_nodes_by_type(network, 'participant')
|
|
initial_supply = np.sum([ network.nodes[i]['holdings'] for i in participants])
|
|
|
|
initial_funds = funds_func(initial_supply)
|
|
|
|
#generate initial proposals
|
|
for ind in range(m):
|
|
j = n+ind
|
|
network.add_node(j)
|
|
network.nodes[j]['type']="proposal"
|
|
network.nodes[j]['conviction']=0
|
|
network.nodes[j]['status']='candidate'
|
|
network.nodes[j]['age']=0
|
|
|
|
r_rv = gamma.rvs(3,loc=0.001, scale=10000)
|
|
network.node[j]['funds_requested'] = r_rv
|
|
|
|
network.nodes[j]['trigger']= trigger_threshold(r_rv, initial_funds, initial_supply)
|
|
|
|
for i in range(n):
|
|
network.add_edge(i, j)
|
|
|
|
rv = np.random.rand()
|
|
a_rv = 1-4*(1-rv)*rv #polarized distribution
|
|
network.edges[(i, j)]['affinity'] = a_rv
|
|
network.edges[(i, j)]['tokens'] = 0
|
|
network.edges[(i, j)]['conviction'] = 0
|
|
network.edges[(i, j)]['type'] = 'support'
|
|
|
|
proposals = get_nodes_by_type(network, 'proposal')
|
|
total_requested = np.sum([ network.nodes[i]['funds_requested'] for i in proposals])
|
|
|
|
network = initial_conflict_network(network, rate = .25)
|
|
network = initial_social_network(network, scale = 1)
|
|
|
|
return network, initial_funds, initial_supply, total_requested
|
|
|
|
def initial_social_network(network, scale = 1, sigmas=3):
|
|
|
|
participants = get_nodes_by_type(network, 'participant')
|
|
|
|
for i in participants:
|
|
for j in participants:
|
|
if not(j==i):
|
|
influence_rv = expon.rvs(loc=0.0, scale=scale)
|
|
if influence_rv > scale+sigmas*scale**2:
|
|
network.add_edge(i,j)
|
|
network.edges[(i,j)]['influence'] = influence_rv
|
|
network.edges[(i,j)]['type'] = 'influence'
|
|
return network
|
|
|
|
def initial_conflict_network(network, rate = .25):
|
|
|
|
proposals = get_nodes_by_type(network, 'proposal')
|
|
|
|
for i in proposals:
|
|
for j in proposals:
|
|
if not(j==i):
|
|
conflict_rv = np.random.rand()
|
|
if conflict_rv < rate :
|
|
network.add_edge(i,j)
|
|
network.edges[(i,j)]['conflict'] = 1-conflict_rv
|
|
network.edges[(i,j)]['type'] = 'conflict'
|
|
return network
|
|
|
|
def social_links(network, participant, scale = 1):
|
|
|
|
participants = get_nodes_by_type(network, 'participant')
|
|
|
|
i = participant
|
|
for j in participants:
|
|
if not(j==i):
|
|
influence_rv = expon.rvs(loc=0.0, scale=scale)
|
|
if influence_rv > scale+scale**2:
|
|
network.add_edge(i,j)
|
|
network.edges[(i,j)]['influence'] = influence_rv
|
|
network.edges[(i,j)]['type'] = 'influence'
|
|
return network
|
|
|
|
def conflict_links(network,proposal ,rate = .25):
|
|
|
|
proposals = get_nodes_by_type(network, 'proposal')
|
|
|
|
i = proposal
|
|
for j in proposals:
|
|
if not(j==i):
|
|
conflict_rv = np.random.rand()
|
|
if conflict_rv < rate :
|
|
network.add_edge(i,j)
|
|
network.edges[(i,j)]['conflict'] = 1-conflict_rv
|
|
network.edges[(i,j)]['type'] = 'conflict'
|
|
return network
|
|
|
|
def social_affinity_booster(network, proposal, participant):
|
|
|
|
participants = get_nodes_by_type(network, 'participant')
|
|
influencers = get_edges_by_type(network, 'influence')
|
|
|
|
j=proposal
|
|
i=participant
|
|
total_inf = np.sum([network.edges[(i,node)]['influence'] for node in participants if (i, node) in influencers ])
|
|
boosts=[network.edges[(node,j)]['affinity']*network.edges[(i,node)]['influence']/total_inf for node in participants if (i, node) in influencers ]
|
|
|
|
return np.sum(boosts)
|
|
|
|
|
|
def trigger_sweep(field, trigger_func,xmax=.2,default_alpha=.5):
|
|
|
|
if field == 'token_supply':
|
|
alpha = default_alpha
|
|
share_of_funds = np.arange(.001,xmax,.001)
|
|
total_supply = np.arange(0,10**9, 10**6)
|
|
demo_data_XY = np.outer(share_of_funds,total_supply)
|
|
|
|
demo_data_Z0=np.empty(demo_data_XY.shape)
|
|
demo_data_Z1=np.empty(demo_data_XY.shape)
|
|
demo_data_Z2=np.empty(demo_data_XY.shape)
|
|
demo_data_Z3=np.empty(demo_data_XY.shape)
|
|
for sof_ind in range(len(share_of_funds)):
|
|
sof = share_of_funds[sof_ind]
|
|
for ts_ind in range(len(total_supply)):
|
|
ts = total_supply[ts_ind]
|
|
tc = ts /(1-alpha)
|
|
trigger = trigger_func(sof, 1, ts)
|
|
demo_data_Z0[sof_ind,ts_ind] = np.log10(trigger)
|
|
demo_data_Z1[sof_ind,ts_ind] = trigger
|
|
demo_data_Z2[sof_ind,ts_ind] = trigger/tc #share of maximum possible conviction
|
|
demo_data_Z3[sof_ind,ts_ind] = np.log10(trigger/tc)
|
|
return {'log10_trigger':demo_data_Z0,
|
|
'trigger':demo_data_Z1,
|
|
'share_of_max_conv': demo_data_Z2,
|
|
'log10_share_of_max_conv':demo_data_Z3,
|
|
'total_supply':total_supply,
|
|
'share_of_funds':share_of_funds}
|
|
elif field == 'alpha':
|
|
alpha = np.arange(.5,1,.01)
|
|
share_of_funds = np.arange(.001,xmax,.001)
|
|
total_supply = 10**9
|
|
demo_data_XY = np.outer(share_of_funds,alpha)
|
|
|
|
demo_data_Z4=np.empty(demo_data_XY.shape)
|
|
demo_data_Z5=np.empty(demo_data_XY.shape)
|
|
demo_data_Z6=np.empty(demo_data_XY.shape)
|
|
demo_data_Z7=np.empty(demo_data_XY.shape)
|
|
for sof_ind in range(len(share_of_funds)):
|
|
sof = share_of_funds[sof_ind]
|
|
for a_ind in range(len(alpha)):
|
|
ts = total_supply
|
|
a = alpha[a_ind]
|
|
tc = ts /(1-a)
|
|
trigger = trigger_func(sof, 1, ts)
|
|
demo_data_Z4[sof_ind,a_ind] = np.log10(trigger)
|
|
demo_data_Z5[sof_ind,a_ind] = trigger
|
|
demo_data_Z6[sof_ind,a_ind] = trigger/tc #share of maximum possible conviction
|
|
demo_data_Z7[sof_ind,a_ind] = np.log10(trigger/tc)
|
|
|
|
return {'log10_trigger':demo_data_Z4,
|
|
'trigger':demo_data_Z5,
|
|
'share_of_max_conv': demo_data_Z6,
|
|
'log10_share_of_max_conv':demo_data_Z7,
|
|
'alpha':alpha,
|
|
'share_of_funds':share_of_funds}
|
|
|
|
else:
|
|
return "invalid field"
|
|
|
|
def trigger_plotter(share_of_funds,Z, color_label,y, ylabel,cmap='jet'):
|
|
dims = (10, 5)
|
|
fig, ax = plt.subplots(figsize=dims)
|
|
|
|
cf = plt.contourf(share_of_funds, y, Z.T, 100, cmap=cmap)
|
|
cbar=plt.colorbar(cf)
|
|
plt.axis([share_of_funds[0], share_of_funds[-1], y[0], y[-1]])
|
|
#ax.set_xscale('log')
|
|
plt.ylabel(ylabel)
|
|
plt.xlabel('Share of Funds Requested')
|
|
plt.title('Trigger Function Map')
|
|
|
|
cbar.ax.set_ylabel(color_label)
|
|
|
|
|
|
def snap_plot(nets, size_scale = 1/500, ani = False, dims = (20,20), savefigs=False):
|
|
|
|
|
|
last_net = nets[-1]
|
|
|
|
last_props=get_nodes_by_type(last_net, 'proposal')
|
|
M = len(last_props)
|
|
last_parts=get_nodes_by_type(last_net, 'participant')
|
|
N = len(last_parts)
|
|
pos = {}
|
|
|
|
for ind in range(N):
|
|
i = last_parts[ind]
|
|
pos[i] = np.array([0, 2*ind-N])
|
|
|
|
for ind in range(M):
|
|
j = last_props[ind]
|
|
pos[j] = np.array([1, 2*N/M *ind-N])
|
|
|
|
if ani:
|
|
figs = []
|
|
fig, ax = plt.subplots(figsize=dims)
|
|
|
|
if savefigs:
|
|
counter = 0
|
|
length = 10
|
|
import string
|
|
unique_id = ''.join([np.random.choice(list(string.ascii_letters + string.digits)) for _ in range(length)])
|
|
for net in nets:
|
|
edges = get_edges_by_type(net, 'support')
|
|
max_tok = np.max([net.edges[e]['tokens'] for e in edges])
|
|
|
|
E = len(edges)
|
|
|
|
net_props = get_nodes_by_type(net, 'proposal')
|
|
net_parts = get_nodes_by_type(net, 'participant')
|
|
net_node_label ={}
|
|
|
|
num_nodes = len([node for node in net.nodes])
|
|
|
|
node_color = np.empty((num_nodes,4))
|
|
node_size = np.empty(num_nodes)
|
|
|
|
edge_color = np.empty((E,4))
|
|
cm = plt.get_cmap('Reds')
|
|
|
|
cNorm = colors.Normalize(vmin=0, vmax=max_tok)
|
|
scalarMap = cmx.ScalarMappable(norm=cNorm, cmap=cm)
|
|
|
|
net_cand = [j for j in net_props if net.nodes[j]['status']=='candidate']
|
|
|
|
for j in net_props:
|
|
node_size[j] = net.nodes[j]['funds_requested']*size_scale/4
|
|
if net.nodes[j]['status']=="candidate":
|
|
node_color[j] = colors.to_rgba('blue')
|
|
trigger = net.nodes[j]['trigger']
|
|
conviction = net.nodes[j]['conviction']
|
|
percent_of_trigger = " "+str(int(100*conviction/trigger))+'%'
|
|
net_node_label[j] = str(percent_of_trigger)
|
|
elif net.nodes[j]['status']=="active":
|
|
node_color[j] = colors.to_rgba('orange')
|
|
net_node_label[j] = ''
|
|
elif net.nodes[j]['status']=="completed":
|
|
node_color[j] = colors.to_rgba('green')
|
|
net_node_label[j] = ''
|
|
elif net.nodes[j]['status']=="failed":
|
|
node_color[j] = colors.to_rgba('gray')
|
|
net_node_label[j] = ''
|
|
elif net.nodes[j]['status']=="killed":
|
|
node_color[j] = colors.to_rgba('black')
|
|
net_node_label[j] = ''
|
|
|
|
for i in net_parts:
|
|
node_size[i] = net.nodes[i]['holdings']*size_scale
|
|
node_color[i] = colors.to_rgba('red')
|
|
net_node_label[i] = ''
|
|
|
|
included_edges = []
|
|
for ind in range(E):
|
|
e = edges[ind]
|
|
tokens = net.edges[e]['tokens']
|
|
edge_color[ind] = scalarMap.to_rgba(tokens)
|
|
if e[1] in net_cand:
|
|
included_edges.append(e)
|
|
|
|
|
|
iE = len(included_edges)
|
|
included_edge_color = np.empty((iE,4))
|
|
for ind in range(iE):
|
|
e = included_edges[ind]
|
|
tokens = net.edges[e]['tokens']
|
|
included_edge_color[ind] = scalarMap.to_rgba(tokens)
|
|
|
|
nx.draw(net,
|
|
pos=pos,
|
|
node_size = node_size,
|
|
node_color = node_color,
|
|
edge_color = included_edge_color,
|
|
edgelist=included_edges,
|
|
labels = net_node_label)
|
|
plt.title('Tokens Staked by Partipants to Proposals')
|
|
|
|
if ani:
|
|
nx.draw(net,
|
|
pos=pos,
|
|
node_size = node_size,
|
|
node_color = node_color,
|
|
edge_color = included_edge_color,
|
|
edgelist=included_edges,
|
|
labels = net_node_label, ax=ax)
|
|
figs.append(fig)
|
|
|
|
else:
|
|
nx.draw(net,
|
|
pos=pos,
|
|
node_size = node_size,
|
|
node_color = node_color,
|
|
edge_color = included_edge_color,
|
|
edgelist=included_edges,
|
|
labels = net_node_label)
|
|
plt.title('Tokens Staked by Partipants to Proposals')
|
|
if savefigs:
|
|
plt.savefig(unique_id+'_fig'+str(counter)+'.png')
|
|
counter = counter+1
|
|
plt.show()
|
|
|
|
if ani:
|
|
False
|
|
#anim = animation.ArtistAnimation(fig, , interval=50, blit=True, repeat_delay=1000)
|
|
#plt.show()
|
|
|