Normed Linear Spaces and Banach Spaces Made Simple

A normed linear space is a vector space with a norm ||x|| measuring each vector's "length". The norm obeys three rules: ||x|| is at least 0 and equals 0 only for the zero vector, ||alpha x|| equals |alpha| times ||x||, and the triangle inequality ||x + y|| <= ||x|| + ||y||. Every norm induces a metric d(x, y) = ||x - y||, so a normed space is automatically a metric space.

Normed spaces are the stage of functional analysis: sequence spaces, function spaces, and solution spaces of differential equations are all normed spaces, and the choice of norm decides which sequences converge. Infinite dimensions bring subtleties: similar-looking norms can give different topologies, and the unit ball is no longer compact.

Key Concepts

• Norm axioms: positivity (||x|| >= 0 with equality only at 0), homogeneity (||alpha x|| = |alpha| ||x||), and the triangle inequality.
• Induced metric: d(x, y) = ||x - y|| turns every normed space into a metric space, so convergence means ||x_n - x|| -> 0.
• Equivalent norms: two norms are equivalent when each is bounded by a constant multiple of the other; equivalent norms give the same convergent sequences and the same topology.
• Examples: R^n with the Euclidean norm; l^p spaces of sequences with (sum |x_n|^p)^{1/p} for 1 <= p < infinity; l^infinity with sup |x_n|; C[0,1] with the sup norm ||f|| = max |f(x)|.
• Unit ball: the set {x : ||x|| <= 1}; in infinite dimensions it is not compact.
• Completeness: a normed space where every Cauchy sequence converges is called a Banach space.

Formulas

• l^p norm: ||x||_p = (|x_1|^p + |x_2|^p + … )^{1/p} for 1 <= p < infinity; for p = 1 this is the sum of absolute values, for p = 2 the square root of the sum of squares.
• Sup norm: ||f||_infinity = sup |f(x)| over the domain; on C[0,1] this is the maximum of |f|, measuring uniform closeness of functions.
• Holder inequality: sum |x_n y_n| <= ||x||_p ||y||_q when 1/p + 1/q = 1; it proves the triangle inequality for l^p norms (Minkowski inequality).
• Scaling: all norms on a finite-dimensional space are equivalent, so R^n has essentially one topology; this fails completely in infinite dimensions.

Worked Example

Check the norm axioms on x = (1, -2, 0, 3, 0, 0, …) in l^1. Its norm is ||x||_1 = 1 + 2 + 0 + 3 = 6. With y = (-1, 1, 0, 0, 0, …), ||y||_1 = 2, and x + y = (0, -1, 0, 3, 0, …) has norm 4, so the triangle inequality holds: 4 <= 6 + 2. Homogeneity: for alpha = -2, ||alpha x||_1 = 2 * 6 = 12, and directly (-2, 4, 0, -6, 0, …) has norm 2 + 4 + 6 = 12. Only the zero sequence has norm 0, confirming positivity.

Common Misconceptions

• "The norm is the same as the absolute value." Absolute value is the norm on R or C; in general a norm is a whole function on a vector space satisfying the three axioms.
• "All norms give the same convergence." True in finite dimensions, false in infinite dimensions: a sequence of functions can converge in one norm and diverge in another.
• "The closed unit ball is compact." In infinite-dimensional normed spaces the closed unit ball is never compact (Riesz's lemma); this is the fundamental difference from R^n.

Real-World Applications

• Signal processing: l^1 and l^2 norms measure signal energy and sparsity; compressed sensing minimizes the l^1 norm to find sparse solutions.
• Approximation theory: the sup norm measures worst-case uniform error of polynomial or neural-network approximations.
• Partial differential equations: Sobolev norms measure functions together with their derivatives and pick the right space for weak solutions.
• Optimization: regularizers like the l^1 norm (lasso) and l^2 norm (ridge) are norms on parameter space shaping the learned model.

A normed space is a vector space with a ruler satisfying positivity, homogeneity, and the triangle inequality. The norm induces the metric and the topology, equivalent norms give the same convergence, and infinite dimensions behave genuinely differently. Test every new space against the three axioms.