Comparative Intelligence and Ethics in Machine Learning Systems

29:15 2 发言者 7 章次章次 573 分部分会议

完整脚本全部搜索

发言者

章次章次

0:00

章章章章章章次第一章章第一章一章 1: Hello, Dr. 304s · Speaker 1

Hello, Dr. Oladako. It's Ben here. I'm here to do a presentation on the machine learning module. And the topic of the presentation is comparative intelligence and ethics in machine learning systems. This is the agenda of my presentation. I …
5:04

章章章章章章次第一章章第一章一章 2: out and prepare the structure activities before the data analysis on both the traditional LM and the deep learning. 294s · Speaker 1

out and prepare the structure activities before the data analysis on both the traditional LM and the deep learning. On the traditional LM, these are the pre -processing activities that were applied, grayscaling, noise reduction, histogram e…
10:00

章章章章章章次第一章章第一章一章 3: the CNN model, and then it used 10 epochs to train the model. 299s · Speaker 2

the CNN model, and then it used 10 epochs to train the model. And when you look at the accuracy, the accuracy is 100 % across the 10 epochs, and then the loss is also zero across the epochs. What does that mean? That indicates that the mode…
15:00

章章章章章章次第一章章第一章一章 4: 337, 505, and 3094 have negative sharp values, meaning that they generally reduce the predicted output rather than increasing it. 299s · Speaker 1

337, 505, and 3094 have negative sharp values, meaning that they generally reduce the predicted output rather than increasing it. Overall, on this visualization, One can say that the SVM model is heavily driven by few key features with thei…
20:00

章章章章章章次第一章章第一章一章 5: and recommend models are always there to validate across subgroups for equal deployment. 83s · Speaker 1

and recommend models are always there to validate across subgroups for equal deployment. The next slide looks at the model deployment. We are not required to deploy the model per se, but the requirement was that we provide the consideration…
21:23

章章章章章章次第一章章第一章一章 6: then monitoring and scaling is applied, and then this is where you track your matrices to auto -scaling to improve performance, and then also check bias and drifting alerts on the model. 302s · Speaker 2

then monitoring and scaling is applied, and then this is where you track your matrices to auto -scaling to improve performance, and then also check bias and drifting alerts on the model. Once that is done, then based on the output or the re…
26:26

章章章章章章次第一章章第一章一章 7: reliance on few dominant features in SVM. 167s · Speaker 1

reliance on few dominant features in SVM. while GrandChem highlighted how the model worked around the X -ray chest image to detect issues for decision making. So this provided transparent and clinical trust to the results. Ethical and bias …