7:00 – 4:30 ASRC NASA/PhD
7:00 – 5:30 ASRC NASA/PhD
7:00 – 4:30 ASRC NASA/PhD
Analyzing Discourse and Text Complexity for Learning and Collaborating
Author: Mihai Dascalu
Notes
7:00 – 4:00 ASRC PhD/NASA
7:00 – 4:30 ASRC PhD/NASA
7:00 – 8:00 (13 hrs) ASRC NASA/PhD
7:00 – 6:00 ASRC PhD
7:00 – 3:00 ASRC NASA
import numpy as np
val = np.array([[0.6]])
row = np.array([[-0.59, 0.75, -0.94,0.34 ]])
col = np.array([[-0.59], [ 0.75], [-0.94], [ 0.34]])
print ("np.dot({}, {}) = {}".format(val, row, np.dot(val, row)))
print ("np.dot({}, {}) = {}".format(col, val, np.dot(col, val)))
'''
note the very different results:
np.dot([[0.6]], [[-0.59 0.75 -0.94 0.34]]) = [[-0.354 0.45 -0.564 0.204]]
np.dot([[-0.59], [ 0.75], [-0.94], [ 0.34]], [[0.6]]) = [[-0.354], [ 0.45 ], [-0.564], [ 0.204]]
'''
# Multiply the values of the relu'd layer [[0, 0.517, 0, 0]] by the goal-output_layer [.61] weight_mat = np.dot(layer_1_col_array, layer_1_to_output_delta) # e.g. [[0], [0.31], [0], [0]] weights_layer_1_to_output_col_array += alpha * weight_mat # add the scaled deltas in # Multiply the streetlights [[1], [0], [1] times the relu2deriv'd input_to_layer_1_delta [[0, 0.45, 0, 0]] weight_mat = np.dot(input_layer_col_array, input_to_layer_1_delta) # e.g. [[0, 0.45, 0, 0], [0, 0, 0, 0], [0, 0.45, 0, 0]] weights_input_to_layer_1_array += alpha * weight_mat # add the scaled deltas in
7:00 – 4:30 ASRC PhD/NASA
[ "a", "about", "above", "after", "again", "against", "all", "am", "an", "and", "any", "are", "as", "at", "be", "because", "been", "before", "being", "below", "between", "both", "but", "by", "could", "did", "do", "does", "doing", "down", "during", "each", "few", "for", "from", "further", "had", "has", "have", "having", "he", "he'd", "he'll", "he's", "her", "here", "here's", "hers", "herself", "him", "himself", "his", "how", "how's", "i", "i'd", "i'll", "i'm", "i've", "if", "in", "into", "is", "it", "it's", "its", "itself", "let's", "me", "more", "most", "my", "myself", "nor", "of", "on", "once", "only", "or", "other", "ought", "our", "ours", "ourselves", "out", "over", "own", "same", "she", "she'd", "she'll", "she's", "should", "so", "some", "such", "than", "that", "that's", "the", "their", "theirs", "them", "themselves", "then", "there", "there's", "these", "they", "they'd", "they'll", "they're", "they've", "this", "those", "through", "to", "too", "under", "until", "up", "very", "was", "we", "we'd", "we'll", "we're", "we've", "were", "what", "what's", "when", "when's", "where", "where's", "which", "while", "who", "who's", "whom", "why", "why's", "with", "would", "you", "you'd", "you'll", "you're", "you've", "your", "yours", "yourself", "yourselves" ]
7:00 – 5:00 ASRC PhD
SUBJECT: dungeon_master1's introduction to the dungeon POST: dungeon_master1 says that you are about to take on a 3-room linear dungeon. SUBJECT: dungeon_master1's introduction to room_0 POST: dungeon_master1 says, The party now finds itself in room_0. There is a troll here. SUBJECT: Asra_Rogueplayer's move in room_0 POST: Asra_Rogueplayer runs from the troll in room_0. SUBJECT: Ping_Clericplayer's move in room_0 POST: Ping_Clericplayer walks towards the troll in room_0. SUBJECT: Valen_Fighterplayer's move in room_0 POST: Valen_Fighterplayer reasons with the troll in room_0. SUBJECT: Emmi_MonkPlayer's move in room_0 POST: Emmi_MonkPlayer walks towards the troll in room_0. SUBJECT: Avia_Bardplayer's move in room_0 POST: Avia_Bardplayer casts a spell at the troll in room_0. SUBJECT: Mirek_Thiefplayer's move in room_0 POST: Mirek_Thiefplayer casts a spell at the troll in room_0. SUBJECT: Lino_Magicplayer's move in room_0 POST: Lino_Magicplayer casts a spell at the troll in room_0. SUBJECT: dungeon_master1's conclusion for room_0 POST: dungeon_master1 says that you have triumphed in the challenge of room_0. SUBJECT: dungeon_master1's introduction to room_1 POST: dungeon_master1 says, The party now finds itself in room_1. There is an idol here. SUBJECT: Asra_Rogueplayer's move in room_1 POST: Asra_Rogueplayer knocks out the idol in room_1. SUBJECT: Ping_Clericplayer's move in room_1 POST: Ping_Clericplayer walks towards the idol in room_1. SUBJECT: Valen_Fighterplayer's move in room_1 POST: Valen_Fighterplayer casts a spell at the idol in room_1. SUBJECT: Emmi_MonkPlayer's move in room_1 POST: Emmi_MonkPlayer examines the idol in room_1. SUBJECT: Avia_Bardplayer's move in room_1 POST: Avia_Bardplayer sneaks by the idol in room_1. SUBJECT: Mirek_Thiefplayer's move in room_1 POST: Mirek_Thiefplayer sneaks by the idol in room_1. SUBJECT: Lino_Magicplayer's move in room_1 POST: Lino_Magicplayer runs from the idol in room_1. SUBJECT: dungeon_master1's conclusion for room_1 POST: dungeon_master1 says that you have triumphed in the challenge of room_1. SUBJECT: dungeon_master1's introduction to room_2 POST: dungeon_master1 says, The party now finds itself in room_2. There is an orc here. SUBJECT: Asra_Rogueplayer's move in room_2 POST: Asra_Rogueplayer casts a spell at the orc in room_2. SUBJECT: Ping_Clericplayer's move in room_2 POST: Ping_Clericplayer reasons with the orc in room_2. SUBJECT: Valen_Fighterplayer's move in room_2 POST: Valen_Fighterplayer knocks out the orc in room_2. SUBJECT: Emmi_MonkPlayer's move in room_2 POST: Emmi_MonkPlayer runs from the orc in room_2. SUBJECT: Avia_Bardplayer's move in room_2 POST: Avia_Bardplayer walks towards the orc in room_2. SUBJECT: Mirek_Thiefplayer's move in room_2 POST: Mirek_Thiefplayer distracts the orc in room_2. SUBJECT: Lino_Magicplayer's move in room_2 POST: Lino_Magicplayer examines the orc in room_2. SUBJECT: dungeon_master1's conclusion for room_2 POST: dungeon_master1 says that you have triumphed in the challenge of room_2. SUBJECT: dungeon_master1's conclusion POST: dungeon_master1 says that you have triumphed in the challenge of the 3-room linear dungeon.
Semantics-Space-Time Cube. A Conceptual Framework for Systematic Analysis of Texts in Space and Time
Listening to How CRISPR Gene Editing Is Changing the World, where Jennifer Kahn discusses the concept of Fitness Cost, where mutations (CRISPR or otherwise) often decrease the fitness of the modified organism. I’m thinking that this relates to the conflicting fitness mechanisms of diverse and monolithic systems. Diverse systems are resilient in the long run. Monolithic systems are effective in the short run. That stochastic interaction between those two time scales is what makes the problem of authoritarianism so hard.
Fitness cost is explicitly modeled here: Kinship, reciprocity and synergism in the evolution of social behaviour
The spread of low-credibility content by social bots
Using Machine Learning to map the field of Collective Intelligence research 
Working on 810 meta-reviews today. Done-ish!
7:00 – 4:00 ASRC PhD/NASA
# based on https://github.com/iamtrask/Grokking-Deep-Learning/blob/master/Chapter6%20-%20Intro%20to%20Backpropagation%20-%20Building%20Your%20First%20DEEP%20Neural%20Network.ipynb
import numpy as np
import matplotlib.pyplot as plt
import typing
# methods --------------------------------------------
# sets all negative numbers to zero
def relu(x: np.array) -> np.array :
return (x > 0) * x
def relu2deriv(output: float) -> float:
return output > 0 # returns 1 for input > 0
# return 0 otherwise
def nparray_to_list(vals: np.array) -> typing.List[float]:
data = []
for x in np.nditer(vals):
data.append(float(x))
return data
def plot_mat(title: str, var_name: str, fig_num: int, mat: typing.List[float], transpose: bool = False):
f = plt.figure(fig_num)
np_mat = np.array(mat)
if transpose:
np_mat = np_mat.T
plt.plot(np_mat)
names = []
for i in range(len(np_mat)):
names.append("{}[{}]".format(var_name, i))
plt.legend(names)
plt.title(title)
# variables ------------------------------------------
np.random.seed(1)
hidden_size= 4
alpha = 0.2
weights_input_to_1_array = 2 * np.random.random((3, hidden_size)) - 1
weights_1_to_output_array = 2 * np.random.random((hidden_size, 1)) - 1
# the samples. Columns are the things we're sampling
streetlights_array = np.array( [[ 1, 0, 1 ],
[ 0, 1, 1 ],
[ 0, 0, 1 ],
[ 1, 1, 1 ] ] )
# The data set we want to map to. Each entry in the array matches the corresponding streetlights_array roe
walk_vs_stop_array = np.array([1, 1, 0, 0]).T # and why are we using the transpose here?
error_plot_mat = [] # for drawing plots
weights_l1_to_output_plot_mat = [] # for drawing plots
weights_input_to_l1_plot_mat = [] # for drawing plots
iter = 0
max_iter = 1000
epsilon = 0.001
layer_2_error = 2 * epsilon
while layer_2_error > epsilon:
layer_2_error = 0
for row_index in range(len(streetlights_array)):
# input holds one instance of the data set at a time
input_layer_array = streetlights_array[row_index:row_index + 1]
# layer one holds the results of the NONLINEAR transformation of the input layer's values (multiply by weights and relu)
layer_1_array = relu(np.dot(input_layer_array, weights_input_to_1_array))
# output layer takes the LINEAR transformation of the values in layer one and sums them (mult)
output_layer = np.dot(layer_1_array, weights_1_to_output_array)
# the error is the difference of the output layer and the goal squared
goal = walk_vs_stop_array[row_index:row_index + 1]
layer_2_error += np.sum((output_layer - goal) ** 2)
# compute the amount to adjust the transformation weights for layer one to output
layer_1_to_output_delta = (goal - output_layer)
# compute the amount to adjust the transformation weights for input to layer one
input_to_layer_1_delta= layer_1_to_output_delta.dot(weights_1_to_output_array.T) * relu2deriv(layer_1_array)
#Still need to figure out why the transpose, but this is where we incrementally adjust the weights
l1t_array = layer_1_array.T
ilt_array = input_layer_array.T
weights_1_to_output_array += alpha * l1t_array.dot(layer_1_to_output_delta)
weights_input_to_1_array += alpha * ilt_array.dot(input_to_layer_1_delta)
print("[{}] Error: {:.3f}, L0: {}, L1: {}, L2: {}".format(iter, layer_2_error, input_layer_array, layer_1_array, output_layer))
#print("[{}] Error: {}, Weights: {}".format(iter, total_error, weight_array))
error_plot_mat.append([layer_2_error])
weights_input_to_l1_plot_mat.append(nparray_to_list(weights_input_to_1_array))
weights_l1_to_output_plot_mat.append(nparray_to_list(weights_1_to_output_array))
iter += 1
# stop even if we don't converge
if iter > max_iter:
break
print("\n--------------evaluation")
for row_index in range(len(streetlights_array)):
input_layer_array = streetlights_array[row_index:row_index + 1]
layer_1_array = relu(np.dot(input_layer_array, weights_input_to_1_array))
output_layer = np.dot(layer_1_array, weights_1_to_output_array)
print("{} = {:.3f} vs. {}".format(input_layer_array, float(output_layer), walk_vs_stop_array[row_index]))
# plots ----------------------------------------------
f1 = plt.figure(1)
plt.plot(error_plot_mat)
plt.title("error")
plt.legend(["layer_2_error"])
plot_mat("input to layer 1 weights", "weight", 2, weights_input_to_l1_plot_mat)
plot_mat("layer 1 to output weights", "weight", 3, weights_l1_to_output_plot_mat)
plt.show()
7:00 – 3:30 ASRC PhD/NASA
viztales 
You must be logged in to post a comment.