Perceptrons and Hopfield Networks
Language
Philosophy in and of Cognitive Science
Neural Networks and Image Classification
Machine Learning Basics
100
This common logical operation/problem, unsolvable by a single perceptron, was a reason why perceptron research did not advance as rapidly in the 50's and 60's.
What is the XOR problem?/ What is XOR?
100
This factorization method for large matrices, was used to reduce the dimensionality of a word x document frequency matrix by identifying the key factors of the matrix when performing LSA (latent semantic analysis).
What is Singular Value Decomposition (SVD)?
100
This level of Marr's Tri-Level analysis is concerned with what the system does and why it does it that way.
What is the computational level?
100
This is a deep feed-forward network for image classification inspired by biological processes.
What is a convolutional network?
100
ACT-R and Shakey the robot are examples of this kind of approach to AI.
What is "GOFAI"/symbolic artificial intelligence?
200
A single layer of perceptrons can only compute this kind of function.
What is a linear function?
200
This is the argument for Chomsky's "Universal Grammar" that children acquire language quickly from few examples.
What is the "poverty of stimulus" argument?
200
DOUBLE JEOPARDY: Double your points or lose 200.
This many months with a redditor's AI husband changed her life.
How long is 9 months.
200
These are the considerations we should be making when setting up a neural network to solve a problem. (4 major categories expected)
What are: the learning rate, hidden node number, the configuration of the model(inputs/outputs + type of network) and the training + testing data? (at least 3/4)
200
This is the learning approach where you train a neural network to perform a predictive task by presenting examples and depending on the correctness of the network's responses altering the weights.
What is supervised learning?
300
The w0 (bias node weight) that given the weights w1 = 0 and w2 = -1, produces the following outputs with the following inputs (Reminder w0 = T):
What is w0 = 1?
300
These types of models capture human judgements about word use but do not
interact with the world, do not have goals/
desires, do not intend to communicate to
other agents.
What are Distributional semantic or statistical
language models?
300
If I lack this, I think Sally should look in the box.
What is theory of mind?
300
This is the final output of the following simple neural network (where every perceptron uses a step activation function), with input x = [2,3]:
What is 1?
300
DOUBLE JEOPARDY: Double your points or lose 200.
You are interested in using a supervised machine learning algorithm to determine how features of a coin (it's size, denomination, weight, color etc) influence human predictions of the outcome of a coin flip (whether they predict heads or tails). This is how your dataset should look.
What is a table with many feature columns (representing coin info) and one outcome column (human prediction)?
400
Given the hopfield network below, this is what the network produces given [-1, +1, -1, -1].
What is [-1, +1, -1, -1]? (Why is this the output?)
400
This is when a type of feature manipulation, where all of the features are used, but they transformed into a new, smaller set of more useful features that capture higher level relationships between the original features. (SVD is one example of this)
What is feature reduction?
400
This kind of model is assessed based on how human-like it is in terms of reaction time, accuracy, mistake made etc.
What is a cognitive model?
400
This method of training a neural network, takes the output of a forward pass, then uses a backwards pass to adjust the weights of the neurons in each layer by calculating the derivative of the error.
What is back-propagation?
400
This is a machine learning approach where correct classifications are not provided during training, and where the goal is to determine similarity of examples and group examples together. K-clustering and feature reduction methods are both examples of this kind of learning.
What is unsupervised learning?
500
Given a 3 node hopfield network and an pattern to learn of [+1, -1, -1] this is what the weight matrix will look like.
What is:
[ 0, -1, -1
-1, 0, +1
-1, +1, 0 ]?
500
This is a representation of word meanings as points and differences in meaning as the angle between words.
What is a semantic space?
500
This is a kind of equivalence between an AI model and human performance that is a measure not just of the model's ability to capture the functions and behavioural output of humans, but also the underlying physical implementation of how those functions are implemented in the brain.
What is structural equivalence?
500
Using the Edge Detection Kernel shown below and a ReLu activation function (with a firing threshold w0 = 5) this is the output given the input matrix x where:
x = [ 2, 4, 0
3, 1, 4
1, 0, 2]
What is 2?
500
This is an approach where a model learns the appropriate policy in a given domain through trial and error, with reward and punishment as its feedback.
What is reinforcement learning?