The COMPAS that the course(Data ethics) shows, does it use a neural network or deep learning?
Timestamp would be helpful.
You didn’t answer my question. Does COMPAS use deep learning that uses something at least a neural network with a loss function and both forward and backpropagation?
Do you know [COMPAS]?
Oh I had no idea what that was
On it’s wikipedia page there is a part that says:
Another general criticism of machine-learning based algorithms is since they are data-dependent if the data are biased, the software will likely yield biased results.
The specific methods used can be found in the research paper:
On page 52:
Advanced Analytical Methods for Risk Prediction
COMPAS from its inception moved beyond traditional Burgess additive scaling to incorpo- rate contemporary data analytic and machine learning (ML) methods for both predictive risk models and for its explanatory/treatment offender classification methods. Specifically for risk prediction models, logistic regression, Lasso (least absolute shrinkage and selection opera- tor) regression, and survival analysis were used for variable selection and weighting. From this developmental work, the final COMPAS risk models for general recidivism and violent recidi- vism were finalized and are documented in Brennan, Dieterich, and Ehret (2009). To develop an “Internal Prison Classification” to guide treatment and management of incarcerated offend- ers bootstrapped K-means methods were used. To correctly assign new intake detainees to this explanatory classification, Random Forests (RF) and Support Vector Machines (SVM) were tested for accuracy of assignment, with RF finally chosen to operationalize the classification software (Breitenbach, Dieterich, & Brennan, 2009; Brennan et al., 2008; Oliver, Brennan, & Dieterich, 2014).
Yes, it was talked on part 1 course 6 video.
I am writing an essay and speech for my speech class that talks about its bias problem sending people to the jail. However, I am just not sure that it uses a deep learning model or not.
The author of this algorithm said it does not use race as a varible, but we all know deep learning model’s bias problem can cause this. Thus, I want to know does it use a deep learning model or not?
No I don’t think they use deep learning models.
Do you know how how COMPAS works?
I read some from the link that I sent…
that is all I know about how it works.