Really impressed with the clarity of the thoughts in the book - The Master Algorithm -How the quest for the ultimate learning machine will remake our world by Pedro Domingos
Notes from the book: (partially)
Symbolists --> learning as the inverse of deduction and take ideas from philosophy, psychology and logic --> master algorithm --> inverse deduction
Connectionists --> reverse engineer the brain and are inspired by neuroscience and physics --> master algorithm --> backpropagation
Evolutionaries --> simulate evolution on the computer and draw on genetics and evolutionary biology --> master algorithm --> genetic programming
Bayesians --> believe learning is a form of probabilistic inference and have their roots in statistics --> master algorithm --> Bayesian inference
Analogizers --> learn by extrapolating from similarity judgements and are influenced by psychology and mathematical optimization --> master algorithm --> support vector machine
* Every algorithm, no matter how complex, can be reduced to just three operations: AND, OR, and NOT. Simple algorithms can be represented by diagrams, using different symbols for the AND, OR and NOT operations.
* an algorithm is not just any set of instructions: they have to be precise and unambigious enough to be executed by a computer.
* algorithms are an exacting standard. Equations are really just a special kind of algorithm.
* designing an algorithm is not easy. pitfalls. find every error and fix it
* complexity such as space, time, and human complexity (intricate for human brains to understand),
In learners world, compiters can write their own programs (machine learning), even so, humans have an ability to control it.
* machine learning is like: farming, learning algorithms are the seeds, datat is the soil, and the learned programs are the grown plants. The machine-learning expert is like a farmer, sowing the seeds, irrigating and fertilizing the soil and keeping an eye of the health of the crop but otherwise staying out of the way.
* some learners learn knowledge and some learn skills.
* machine learning takes many different forms and goes by many different names: pattern recognition, statistical modeling, data mining, knowledge discovery, predictive analytics, data science, adaptive systems, self-organizing systems,
machine learning is a subfield of AI. Goal of AI is to teach computers to do what humans currently do better and learning is arguably the most important of those things.
statistically correct is required, you cannot have deterministically correct solutions all the time.
** The industrial revolution automated manual work and the information revoluton did the same for mental work, but machine learning automates automation itself.
learning alorithms are the matchmakers: they find producers and consumers for each other, curring through the information overload.
*** the best way for a company to ensure that learners likes its products is to run them itself. whoever has the best algorithms and the most data wins. a new type of network effect takes hold: whoever has the most customers accumulates the most data, learns the best models, wins the most new customers, and so on in a virtuous circle.
** "data is the new oil" you have to refine it properly."
* machine learning is the scientific method on steroids, it follows the same process of generating, testing and discarding or refining hypotheses.
The biggest challenge is assembling all this information into a coherent whole.
once a learned program is deployed, the bad guys change their behavious to defeat it. This constrasts with the natural world, which always works the same way. the solution is to marry machine learning with game theory, don't just learn to defeat what your opponent does now; learn to parry what he might do against the learner. factoring in the costs and benefits of different actions, as game theory does, can also help strike the right balance between privacy and security.
Chapter 2: The Master Algorithm / Pg23 (46/353)
The machine learning algorithm must deploy same algorithm for doing all of the different things (not write two different programs). For example, the same ML should help play chess as well as do the credit-card applications processing.
Bayes equations -- diagnose the condition based on the database of patient records - their symptoms, test results. The same algorithm is used to learn spam filters.
Nearest-neighbor algo -- hand-writing recogniton, controlling robot hands, recommending books and movies
Decision tree learners - credit card application acceptance, splice junctions in DNA, choosing the next move in a game of chess.
enough data could be infinite. Learning from finite data requires making assumptions, and different learners make different assumptions, which makes them good for some things but not others.
All knowledge - past, present, and future - can be derived from data by a single, universal learning algorithm ==> Master Algorithm, this will be the last thing we will every have to invent because once we let it loose, it will go on to invent everything else that can be invented.
Evolution is an iterative search, solving a problem by trying many candidate solutions, selecting and modifying the best ones, and repeating these steps as many times as necessary.
Notes from the book: (partially)
Symbolists --> learning as the inverse of deduction and take ideas from philosophy, psychology and logic --> master algorithm --> inverse deduction
Connectionists --> reverse engineer the brain and are inspired by neuroscience and physics --> master algorithm --> backpropagation
Evolutionaries --> simulate evolution on the computer and draw on genetics and evolutionary biology --> master algorithm --> genetic programming
Bayesians --> believe learning is a form of probabilistic inference and have their roots in statistics --> master algorithm --> Bayesian inference
Analogizers --> learn by extrapolating from similarity judgements and are influenced by psychology and mathematical optimization --> master algorithm --> support vector machine
* Every algorithm, no matter how complex, can be reduced to just three operations: AND, OR, and NOT. Simple algorithms can be represented by diagrams, using different symbols for the AND, OR and NOT operations.
* an algorithm is not just any set of instructions: they have to be precise and unambigious enough to be executed by a computer.
* algorithms are an exacting standard. Equations are really just a special kind of algorithm.
* designing an algorithm is not easy. pitfalls. find every error and fix it
* complexity such as space, time, and human complexity (intricate for human brains to understand),
In learners world, compiters can write their own programs (machine learning), even so, humans have an ability to control it.
* machine learning is like: farming, learning algorithms are the seeds, datat is the soil, and the learned programs are the grown plants. The machine-learning expert is like a farmer, sowing the seeds, irrigating and fertilizing the soil and keeping an eye of the health of the crop but otherwise staying out of the way.
* some learners learn knowledge and some learn skills.
* machine learning takes many different forms and goes by many different names: pattern recognition, statistical modeling, data mining, knowledge discovery, predictive analytics, data science, adaptive systems, self-organizing systems,
machine learning is a subfield of AI. Goal of AI is to teach computers to do what humans currently do better and learning is arguably the most important of those things.
statistically correct is required, you cannot have deterministically correct solutions all the time.
** The industrial revolution automated manual work and the information revoluton did the same for mental work, but machine learning automates automation itself.
learning alorithms are the matchmakers: they find producers and consumers for each other, curring through the information overload.
*** the best way for a company to ensure that learners likes its products is to run them itself. whoever has the best algorithms and the most data wins. a new type of network effect takes hold: whoever has the most customers accumulates the most data, learns the best models, wins the most new customers, and so on in a virtuous circle.
** "data is the new oil" you have to refine it properly."
* machine learning is the scientific method on steroids, it follows the same process of generating, testing and discarding or refining hypotheses.
The biggest challenge is assembling all this information into a coherent whole.
once a learned program is deployed, the bad guys change their behavious to defeat it. This constrasts with the natural world, which always works the same way. the solution is to marry machine learning with game theory, don't just learn to defeat what your opponent does now; learn to parry what he might do against the learner. factoring in the costs and benefits of different actions, as game theory does, can also help strike the right balance between privacy and security.
Chapter 2: The Master Algorithm / Pg23 (46/353)
The machine learning algorithm must deploy same algorithm for doing all of the different things (not write two different programs). For example, the same ML should help play chess as well as do the credit-card applications processing.
Bayes equations -- diagnose the condition based on the database of patient records - their symptoms, test results. The same algorithm is used to learn spam filters.
Nearest-neighbor algo -- hand-writing recogniton, controlling robot hands, recommending books and movies
Decision tree learners - credit card application acceptance, splice junctions in DNA, choosing the next move in a game of chess.
enough data could be infinite. Learning from finite data requires making assumptions, and different learners make different assumptions, which makes them good for some things but not others.
All knowledge - past, present, and future - can be derived from data by a single, universal learning algorithm ==> Master Algorithm, this will be the last thing we will every have to invent because once we let it loose, it will go on to invent everything else that can be invented.
Evolution is an iterative search, solving a problem by trying many candidate solutions, selecting and modifying the best ones, and repeating these steps as many times as necessary.
No comments:
Post a Comment