Algorithms to Live By

• Optimal stopping problem
• Explore vs Exploit - Explore is when you're researching / trying something new / breaking new ground (but not reaping any tangible value). Exploit on the other hand is when you begin draining value from the aforementioned explore phase. An example of this is movie studios who churn out sequels (as known-ish guarantees of success).
• However note that exploitation cannot last forever. It'll eventually exhaust. Explore vs Exploit is something continually faced by drug companies.
• "One of the deepest truths about machine learning is that it's not always better to use a more complex model; one that takes a greater number of factors into account [as it can lead to overfitting]"
• Overfitting occurs when there is noise or mismeasurement in our data (and there almost always is). If the model is too complex it will find patterns in the data, even when they are mere noise.
• Overfitting is a consequence of the idolatry of data.
• Power law, normal, and erlang
• "The further ahead you need to brainstorm the thicker the brush you should use"
• Overfitting / Regularization (e.g. Thinking and doing less / it can be a waste of time and effort and can lead to a worse solution ... consider "early stopping")
• Bloom filters / Monte Carlo
• Local vs global maximum / Hills, valleys, and traps, Hill climbing algorithm, considering factoring some of this into future trips and travel ... leave ATX for a week with no route / destination (other than to come home)
• Linguistic Data / University of Penn
• Metropolis Algorithm, Scatter Gun Algorithm, Jitter / Random Restarts
• Randomly jitter your life / Wikipedia random entry Exponential Backoff
• Vickrey auction and blockchain / always be "computationally kind" (suggest dates when scheduling)