Add random search baseline

Generate a graph of the performance of a random search
for a given kernel and inputs sizes.

Author: Jules Pondard

python/experimental/options_search/random_search.py

@@ -0,0 +1,63 @@
+import numpy as np
+#import ipdb
+import torch
+import torch.optim as optim
+import torch.nn as nn
+import torch.utils.data
+import torch.nn.functional as F
+import tensor_comprehensions as tc
+from visdom import Visdom
+
+import utils
+
+NB_EPOCHS = 1000
+BATCH_SZ = 1
+
+viz = Visdom()
+win0 = viz.line(X=np.arange(NB_EPOCHS), Y=np.random.rand(NB_EPOCHS))
+
+(tc_code, tc_name, inp, init_input_sz) = utils.get_convolution_example()
+
+NB_HYPERPARAMS, INIT_INPUT_SZ = utils.NB_HYPERPARAMS, utils.INIT_INPUT_SZ
+
+def getRandom():
+    opt_v = np.zeros(NB_HYPERPARAMS).astype(int)
+    for i in range(opt_v.shape[0]):
+        opt_v[i] = np.random.randint(utils.cat_sz[i])
+    return opt_v
+
+
+INTER_DISP = 20
+
+running_reward = -0.5
+tab_rewards=[]
+tab_best=[]
+best=-12
+best_options = -1
+for i in range(NB_EPOCHS):
+    rewards = []
+    opts=[]
+    for j in range(BATCH_SZ):
+        out = getRandom()
+        reward = utils.evalTime(out.astype(int), prune=2, curr_best=np.exp(-best))
+        reward = -np.log(reward)
+        rewards.append(reward)
+        opts.append(out.astype(int))
+    if(best < np.max(rewards) or i==0):
+        best = np.max(rewards)
+        ind=np.argmax(rewards)
+        best_options = opts[ind]
+        utils.print_opt(best_options)
+    if(i==0):
+        running_reward = reward
+    running_reward = running_reward * 0.99 + np.mean(rewards) * 0.01
+    tab_rewards.append(-running_reward)
+    tab_best.append(-best)
+    if i % INTER_DISP == 0:
+        viz.line(X=np.column_stack((np.arange(i+1), np.arange(i+1))), Y=np.column_stack((np.array(tab_rewards), np.array(tab_best))), win=win0, opts=dict(legend=["Geometric run", "Best time"]))
+    print(-running_reward)
+    print(-best)
+tab_best = np.array(tab_best)
+np.save("randomsearch.npy", tab_best)
+print("Finally, best options are:")
+utils.print_opt(best_options)