A running post of small things that I learn everyday!
2024
2nd Oct
- Identifying communities in networks using top-down and bottom up approaches
- Modularity metric for communities
- Using the right evaluation metrics for class imbalance problems are important. Precision, Recall and F1 are asymmetric metrics which means their values change depending on which class is considered positive class
1st Oct
- Biasing methods in word embeddings and their potential risks.
- Bias in word embeddings can lead to potentially amplification of existing societal stereotypes. This is extremely harmful. Removing such biases however can also lead to loss of context in model. Also, biases are quite subjective, so which biases are removed and which aren’t are a big question overall.
29th Sept
27th Sept
- Learnt how to handle custom domains on google websites
- Google indexing on your profiles will work much better if your name and social media user name match (for eg, my insta handle is _kushaaaaaa , had it been kushasahu which was already taken, it would have ranked higher)
- Small world property of a network can be analysed by checking its clustering coefficient with similar random networks and ASPL with corresponding random networks. Clustering coefficient should be higher than random networks and ASPL should be statistically diff but within 10 fold.
24th Sept