def generatePlot(ax, exp_paths, bounds):
for exp_path in exp_paths:
exp = ExperimentModel.load(exp_path)
results = loadResults(exp, 'return_summary.npy')
plotSensitivity(results, param, ax, label=exp.agent)
def generatePlotTTA(ax, exp_path, bounds):
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
const, unconst = tee(results)
color = colors[exp.agent]
label = rename(exp.agent)
if 'ReghTDC' in exp.agent:
const = whereParameterEquals(const, 'ratio', 1)
const = whereParameterEquals(const, 'reg_h', 1)
elif 'TDRCC' in exp.agent:
const = whereParameterEquals(const, 'ratio', 1)
# const = whereParameterEquals(const, 'reg_h', 0.8)
const = whereParameterGreaterEq(const, 'reg_h', 0.01)
elif 'TDC' in exp.agent:
const = whereParameterGreaterEq(const, 'ratio', 1)
if show_unconst:
b = plotSensitivity(unconst, param, ax, stderr=stderr, color=color, label=label + '_unc', bestBy=bestBy, dashed=True)
bounds.append(b)
b = plotSensitivity(const, param, ax, stderr=stderr, color=color, label=label, bestBy=bestBy)
bounds.append(b)
def generatePlot(exp_paths):
ax = plt.gca()
bounds = []
for exp_path in exp_paths:
exp = loadExperiment(exp_path)
results = loadResults(exp)
if exp.agent == 'TDadagrad':
continue
use_ideal_h = exp._d['metaParameters'].get('use_ideal_h', False)
dashed = use_ideal_h
color = colors[exp.agent]
label = exp.agent.replace('adagrad', '')
if use_ideal_h:
label += '-h*'
bound = plotSensitivity(results, 'h_variance', ax, label=label, color=color, dashed=dashed, bestBy='end')
bounds.append(bound)
lower = min(map(lambda x: x[0], bounds))
upper = max(map(lambda x: x[1], bounds))
ax.set_ylim([lower, upper])
ax.set_xscale("log", basex=2)
# plt.show()
save(exp, f'h_variance-sensitivity')
plt.clf()
def generatePlotTTA(ax, exp_paths, bestBy, bounds):
ax.set_xscale("log", basex=2)
for exp_path in exp_paths:
if 'amsgrad' in exp_path:
continue
exp = loadExperiment(exp_path)
rmsve = loadResults(exp, 'errors_summary.npy')
rmspbe = loadResults(exp, 'rmspbe_summary.npy')
agent = exp.agent
if 'SmoothTDC' in agent:
average = exp._d['metaParameters']['averageType']
agent += '_' + average
color = colors[agent]
label = agent
b = plotSensitivity(rmsve,
'ratio',
ax,
overStream=rmspbe,
color=color,
label=label,
bestBy=bestBy)
bounds.append(b)
def generatePlot(exp_paths):
ax = plt.gca()
bounds = []
i = -1
for exp_path in exp_paths:
i += 1
exp = loadExperiment(exp_path)
results = loadResults(exp, 'errors_summary.npy')
param_dict = splitOverParameter(results, 'ratio')
for key in param_dict:
bound = plotSensitivity(param_dict[key],
'alpha',
ax,
color=colors[i],
bestBy='end')
bounds.append(bound)
lower = min(map(lambda x: x[0], bounds))
upper = max(map(lambda x: x[1], bounds))
ax.set_ylim([lower, upper])
ax.set_xscale("log", basex=2)
# plt.show()
save(exp, f'alpha_over_beta_rmsve', type='svg')
plt.clf()
def generatePlot(exp_paths):
ax = plt.gca()
bounds = []
for exp_path in exp_paths:
exp = loadExperiment(exp_path)
results = loadResults(exp)
if exp.agent == 'TDadagrad':
continue
color = colors[exp.agent]
label = exp.agent.replace('adagrad', '')
bound = plotSensitivity(results, 'h_variance', ax, label=label, color=color, bestBy='end')
bounds.append(bound)
# lower = min(map(lambda x: x[0], bounds))
# upper = max(map(lambda x: x[1], bounds))
# ax.set_ylim([lower, upper])
ax.set_xscale("log", basex=2)
plt.show()
# save(exp, f'alpha-sensitivity')
plt.clf()
def generatePlot(ax, exp_path, bounds):
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
color = colors[exp.agent]
label = exp.agent
b = plotSensitivity(results, param, ax, color=color, label=label, bestBy=bestBy)
bounds.append(b)
def generatePlot(ax, exp_path, bounds):
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
results = addUpdateParam(results)
color = colors[exp.agent]
label = rename(exp.agent)
b = plotSensitivity(results,
"updates",
ax,
color=color,
label=label,
bestBy=bestBy)
bounds.append(b)
def generatePlotTTA(ax, exp_paths, bestBy, bounds):
ax.set_xscale("log", basex=2)
for exp_path in exp_paths:
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
agent = exp.agent
color = colors[agent]
label = agent
b = plotSensitivity(results,
'reg_h',
ax,
reducer='best',
color=color,
label=label,
bestBy=bestBy)
bounds.append(b)
def generatePlotTTA(ax, exp_paths, bestBy, bounds):
ax.set_xscale("log", basex=2)
for exp_path in exp_paths:
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
agent = exp.agent
if 'SmoothTDC' in agent:
average = exp._d['metaParameters']['averageType']
agent += '_' + average
color = colors[agent]
label = agent
b = plotSensitivity(results,
'buffer',
ax,
color=color,
label=label,
bestBy=bestBy)
bounds.append(b)
def generatePlotTTA(ax, exp_paths, bestBy, bounds):
ax.set_xscale("log", basex=2)
for exp_path in exp_paths:
exp = loadExperiment(exp_path)
results = loadResults(exp, errorfile)
agent = exp.agent
color = colors[agent]
label = agent
if exp.agent == 'ReghTDC':
results = whereParameterGreaterEq(results, 'ratio', 1.0)
# reducer='best' chooses the best value of other parameters *per value of 'replay'*
# reducer='slice' first chooses the best parameter setting, then sweeps over 'replay' with other parameters fixed
b = plotSensitivity(results,
'replay',
ax,
reducer='best',
color=color,
label=label,
bestBy=bestBy)
bounds.append(b)