Information theory is a conceptual framework quantifying information, entropy, and data compression in physical and biological systems.The central concept is Shannon's information entropy: H = −∑ pᵢ log₂ pᵢ. It measures the average number of bits needed to encode one source symbol. If all symbols are equally likely, H is maximized. The redundancy of language allows lossless data compression (Shannon proved the compression limit in the source coding theorem). Another cornerstone is channel capacity: C = B log₂(1+S/N). If you encode a message at a rate lower than C, you can transmit it practically without errors, despite noise. These ideas work in all communication systems: from satellites to USB flash drives.
Claude Shannon (1916–2001) formulated the mathematical theory in his secret wartime reports of the early 1940s and summarized it in the famous 1948 paper in the Bell System Technical Journal. He introduced the term 'bit' (suggested by J. Tukey) and linked information entropy with Boltzmann's physical entropy. Predecessors included H. Nyquist (studied telegraph transmission speed, 1924) and R. Hartley (proposed a logarithmic measure, 1928). Later, A.N. Kolmogorov developed an algorithmic approach to information (Kolmogorov complexity). In quantum information theory, von Neumann entropy generalizes Shannon entropy for mixed quantum states.
Information is measured in bits as the logarithm of the number of possible states: I = log₂ N or −log₂ p. Entropy H = −∑ pᵢ log pᵢ reaches its maximum with a uniform distribution. When transmitted over a channel with capacity C (Shannon-Hartley formula: C = B log₂(1+S/N) for Gaussian noise), Shannon's theorem guarantees: if the coding rate R < C, then an arbitrarily low error probability can be achieved using sufficiently long block codes. In practice, the limit is approached using turbo and LDPC codes, which employ iterative decoding and prior probabilities. For lossless compression, the optimal code achieves an average length equal to entropy and is constructed using Huffman's algorithm or arithmetic coding. Information theory is also applied in machine learning (minimum description length principle, Bayesian methods) and bioinformatics.
Thanks to it, the internet, JPEG photo compression, and the recovery of scratched music on CDs work. Even DNA sequencing uses an information measure to assemble the genome.
The phrase 'Hello, World!' in random English text would contain about 80 bits, but due to language redundancy it can be compressed to 40-50 bits. Shannon calculated that English text carries on average only 1.3 bits per letter.
| I | variable | information flow between subsystems — information flow bit per second |
| σ | variable | entropy production rate — entropy production rate joule per kelvin second |
| A | variable | activity (a measure of dynamical activity in the system) — activity per second |
| β | variable | inverse temperature (1/k_B T) — inverse temperature per joule |
| |·|² | operator | squared magnitude — absolute value squared |
| d/dt | operator | derivative with respect to time — time derivative |


















