img:hover { transform: scale(1.05); } Today I fine tuned an image recognition deep learning model, created a web interface with Gradio, and deployed it on Hugging Face Spaces. This project was the subject of lesson 2 of fast.ai’s Deep Learning for Coders course. 1. Fine-tuning a Model in Kaggle Deep learning models need to train on a GPU (Graphics Processing Unit) because CPUs are too slow. I followed fast.
Interpreting the coefficient estimators in a logistic regression is straightforward. The binary logistic regression model is \[y = \mathrm{logit}(\pi) = \ln\left(\frac{\pi}{1 - \pi}\right) = X\beta\] A \(\delta = x_1 - x_0\) unit change in \(x\) in a estimated regression \(\hat{y} = X\hat{\beta}\) is associated with a \(\delta\hat{\beta}\) factor change in the log odds of \(y\). More commonly, you take the exponent of the coefficient estimate and say a \(\delta\) unit change is associated with a \(e^{\delta\hat{\beta}}\) factor change in the odds of \(y\) (see my handbook).
Buffer published The 2021 State of Remote Work, a summary of a survey of over 2,000 remote workers in late 2020 and made the data available as a Google sheets file on Google Docs. BTW, the report is a follow up to their 2020 survey which was featured in MakeoverMonday, a great program devoted to rethinking data visualization. Buffer found that remote workers overwhelmingly support remote work. However, they did share struggles such as difficulty unplugging from the office, collaborating, and loneliness.