2024-04-30
- Lottery ticket hypothesis: take a full model and prune it down to get the same performance. Reweight. Rebalance. Simplify. Iterate. The multiple things.
2024-04-29
- when doing image classification you 1) label the data and 2) create batches of support sets and query sets, 3) training episodes where you update and reweight the embedding a best classify the query set classes.
2024-04-18
Chapter 2
- you can unsupervise learn something by adding noise to the truth and minimizing the distance in the embeddings between truth and noisy truth and maximize distance from something else.
- Transfer learning: one general model, then update the target variable and dataset for a smaller model.
2024-04-17
_________________________
Bryan lives somewhere at the intersection of faith, fatherhood, and futurism and writes about tech, books, Christianity, gratitude, and whatever’s on his mind. If you liked reading, perhaps you’ll also like subscribing: