opensource/manim
1
0
Fork 0
mirror of https://github.com/3b1b/manim.git synced 2025-07-29 21:12:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

After running 2to3

Browse Source
This commit is contained in:
Grant Sanderson
2018-08-09 17:56:05 -07:00
parent 06a65190e7
commit 858051a806
172 changed files with 2117 additions and 2221 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
old_projects/nn/network.py
View File

@ -8,7 +8,7 @@ using backpropagation. Note that I have focused on making the code
simple, easily readable, and easily modifiable. It is not optimized,
and omits many desirable features.
"""
from __future__ import print_function
#### Libraries
# Standard library
@ -18,7 +18,7 @@ import random
import numpy as np
import os
from PIL import Image
import cPickle
import pickle
from nn.mnist_loader import load_data_wrapper
NN_DIRECTORY = os.path.dirname(os.path.realpath(__file__))
@ -91,11 +91,11 @@ class Network(object):
tracking progress, but slows things down substantially."""
if test_data: n_test = len(test_data)
n = len(training_data)
for j in xrange(epochs):
for j in range(epochs):
random.shuffle(training_data)
mini_batches = [
training_data[k:k+mini_batch_size]
for k in xrange(0, n, mini_batch_size)]
for k in range(0, n, mini_batch_size)]
for mini_batch in mini_batches:
self.update_mini_batch(mini_batch, eta)
if test_data:
@ -147,7 +147,7 @@ class Network(object):
# second-last layer, and so on. It's a renumbering of the
# scheme in the book, used here to take advantage of the fact
# that Python can use negative indices in lists.
for l in xrange(2, self.num_layers):
for l in range(2, self.num_layers):
z = zs[-l]
sp = self.d_non_linearity(z)
delta = np.dot(self.weights[-l+1].transpose(), delta) * sp
@ -196,7 +196,7 @@ def ReLU_prime(z):
def get_pretrained_network():
data_file = open(PRETRAINED_DATA_FILE)
weights, biases = cPickle.load(data_file)
weights, biases = pickle.load(data_file)
sizes = [w.shape[1] for w in weights]
sizes.append(weights[-1].shape[0])
network = Network(sizes)
@ -210,7 +210,7 @@ def save_pretrained_network(epochs = 30, mini_batch_size = 10, eta = 3.0):
network.SGD(training_data, epochs, mini_batch_size, eta)
weights_and_biases = (network.weights, network.biases)
data_file = open(PRETRAINED_DATA_FILE, mode = 'w')
cPickle.dump(weights_and_biases, data_file)
pickle.dump(weights_and_biases, data_file)
data_file.close()
def test_network():
@ -241,7 +241,7 @@ def maximizing_input(network, layer_index, layer_vect, n_steps = 100, seed_guess
else:
pre_sig_guess = np.random.randn(weights[0].shape[1])
norms = []
for step in xrange(n_steps):
for step in range(n_steps):
activations = network.get_activation_of_all_layers(
sigmoid(pre_sig_guess), layer_index
)
@ -269,19 +269,19 @@ def save_organized_images(n_images_per_number = 10):
training_data, validation_data, test_data = load_data_wrapper()
image_map = dict([(k, []) for k in range(10)])
for im, output_arr in training_data:
if min(map(len, image_map.values())) >= n_images_per_number:
if min(list(map(len, list(image_map.values())))) >= n_images_per_number:
break
value = int(np.argmax(output_arr))
if len(image_map[value]) >= n_images_per_number:
continue
image_map[value].append(im)
data_file = open(IMAGE_MAP_DATA_FILE, mode = 'w')
cPickle.dump(image_map, data_file)
pickle.dump(image_map, data_file)
data_file.close()
def get_organized_images():
data_file = open(IMAGE_MAP_DATA_FILE, mode = 'r')
image_map = cPickle.load(data_file)
image_map = pickle.load(data_file)
data_file.close()
return image_map