By: Hannah Pell Physics
However, sometimes it can be just as helpful to know what we can’t do. Just as there are rules that govern our daily lives limiting what’s permissible — no talking in the library; no running by the pool; no shirt, no shoes, no service — so too there are restrictions on what’s permissible in the physical world. Many of these rules can be grouped together simply as the “no” theorems.
No-Go Theorems and Quantum Information Theory
There are several theorems that are the “no-go” type within quantum information theory. One is the “no-communication” (or “no-signaling”) theorem which shows that instantaneous sharing of information between two observers is an impossibility. This theorem is an important qualifier to quantum entanglement (more popularly known as Einstein’s “spooky action at a distance”). Quantum entanglement is an effect that may characterize widely separated but potentially correlated events, and the “no-communication” theorem restricts the information that can be communicated between two far-away observers of an entangled state.
Remember our cryptographic friends Alice and Bob? The “no-communication” theorem says that if Alice and Bob are two independent observers taking simultaneous measurements on parts of a quantum-mechanical system, there is no action that Alice can perform on her part that would be detectable by Bob while observing his part (and vice versa).
There are also the “no-cloning,” “no-broadcast,” and “no-deleting” theorems. According to the “no-cloning” theorem, pure quantum states — which can be described by a single vector — cannot be identically copied. The “no-broadcast” theorem extends this to mixed quantum states, or combinations of pure quantum states. The “no-deleting” theorem states that if there are two copies of a quantum state, it is impossible to delete one of the copies. The no-cloning and no-deleting theorems together imply that quantum information is conserved — like energy, it cannot be created or destroyed (this is also described in the “no-hiding” theorem).
No Free Lunch Theorem and Machine Learning
The “no free lunch” theorem in machine learning implies something similar. The theorem originated in a 1997 publication titled “No Free Lunch Theorems for Optimization” authored by physicists David H. Wolpert and William G. Macready, who wrote that no free lunch theorems “establish that for any algorithm, any elevated performance over one class of problems is offset by performance over another class.” In other words, Wolpert and Macready showed that the computational cost for finding a solution is the same for any solution method when averaged overall problems in the particular class. Even when the goal is optimization, there are no shortcuts to the desired solution — no free lunch.
There is still no proof of a generalized no-hair theorem, and mathematicians refer to it as the “no-hair conjecture” because of this. A conjecture differs from a theorem in that it is mathematically unproven but thought to be true, whereas a theorem has been proven true. Even though the no-hair theorem lacks rigorous proof, it is in accordance with general relativity, so it’s widely accepted as valid within the physics community.