Comparisons
Methodologies
Terminology
Tricks
Misc (Club/Tools)
100
As opposed to being _____,
_____ data can be broken into finite points.
What are Continuous and Discrete?
100
A measurement of how well our hypothesis fits the data.
What is the Cost Function?
100
This may occur when we have high variance in our data.  (Hint: Basically when we tightly adapt to the noise of the training data)
What is Overfitting?
100
The practice of putting inputs to a common scale.  This technique has been shown to improve convergence speed, and may also improve certain classifiers.
What is Normalization (Feature Scaling)?
100

This Hanson Robotics robot was named an official citizen of Saudi Arabia late last year.

Who is Sophia?

200
_____ would be the proper term for output data, whereas
_____ would be the name for input data.
What are Labels and Features?
200
The means of propelling us from one set of nodes to the next in a network.
What is Forward Propagation?
200
The connections, or weights, between neuron layers.

What are Synapses?

200
This trick has been used on numerous occasions to improve object detection by mapping it within its environment.
What is SLAM?
200

Our dedicated outreach officer, who graciously takes the time to record club moments and craft our weekly newsletter.

Who is Ignacio Granda Lutz?

300
_____ regression produces discrete labels, whereas
_____ regression produces continuous labels.
What are Logistic and Linear?
300
This algorithm estimates how features should scale to improve our hypothesis.  (Hint:  Incorporates the cost function)
What is Gradient Descent?
300

The practice of using many hidden layers, along with greater overall layer connectivity, to achieve more complex models.

What is Deep Learning?

300
This helpful trick mathematically reduces overfitting during training.
What is Regularization?
300
This popular robotics company has built a variety of robots, from Atlas, to Hopper, to Spot Mini.  Though formerly Google's, they are now working under Japan's SoftBank Group.
Who is Boston Dynamics?
400

_____ learning trains a computer through incentivization,
_____ learning trains a computer to categorize without labels.

What are Reinforcement and Unsupervised?
400

This algorithm performs clustering by assigning points to centroids (cluster centers), and adjusting centroids as more points are added.


What is K-Means?

400
In our Computer Vision workshop, we applied these as our way of grabbing outlines in our images and recognizing digits.
What are Contours?
400
By altering its traditional implementation, this machine learning model practices convolution, pooling, and filtering, and is thereby great at recognizing images.
What is a Convolutional Neural Network?
400

A popular Machine Learning library that runs on top of Tensor Flow (or Theano, if you desire).

What is Keras?

500
_____ separates groups by their widest gaps, while
_____ passes features into a system of layered, weighted nodes.
What are Support Vector Machines and Neural Networks?
500
A learning model that branches through a series of rules, until it reaches end-states/predictions.
What are Decision Trees?
500
A full iteration over an entire dataset -- often used in Neural Networks.
What is an Epoch?
500
Used more commonly in SVMs, this shortcut allows our kernel to operate at higher dimensions without computing coordinates.
What is the Kernel Trick?
500
This Google Deepmind project defeated world-renowned Go champion, Lee Sedol, in 2016.
What is AlphaGo?
M
e
n
u