Notation

This chapter is the master notation table for the whole book. The conventions here are field-standard and override the courseware wherever they differ; the “Course writes” column records each deviation so you can cross-reference Zhang Bo’s slides. The choice of notation policy is recorded in docs/adr/0002-pragmatic-notation-standardization.md, and the canonical word for each concept in CONTEXT.md.

Sets, spaces, and general symbols

Symbol Meaning Course writes
\(\mathbb{R}, \mathbb{R}^n, \mathbb{R}_+\) reals, \(n\)-space, non-negative reals same
\(X \subseteq \mathbb{R}^n\) feasible set / admissible decisions \(\aleph\)
\(f:\mathbb{R}^n\to\mathbb{R}\) static objective \(F\), with codomain \(\Im\)
\(\nabla f,\ \nabla^2 f\) gradient, Hessian same
\(\langle a,b\rangle,\ a^\top b\) inner product same
\(a \odot b\) componentwise (Hadamard) product \(a\circ b\)
\(:=\) “is defined as” \(\triangleq\) / \(=\)

Static optimization (Ch. 1)

Symbol Meaning Course writes
\(g(x)\ge 0\) inequality constraints (sign convention kept) same (\(g\ge 0\))
\(h(x)=0\) equality constraints same
\(\lambda \ge 0\) KKT multipliers on \(g\) \(\lambda\)
\(\mu\) multipliers on \(h\) \(\mu\)
\(\lambda_0\) Fritz–John multiplier on the objective \(\lambda_0\)
\(\mathcal{L}(x,\lambda,\mu)\) Lagrangian \(L\)
\(V(\theta)\) value function of a parameter \(\theta\) (envelope) same

Dynamics and ODEs (Ch. 2)

Symbol Meaning Course writes
\(\dot x = \mathrm{d}x/\mathrm{d}t\) time derivative same (also \(x'\))
\(A\), \(\lambda_i\), \(\xi_i\) system matrix, eigenvalues, eigenvectors same
\(\tau=\operatorname{tr}A,\ \Delta=\det A\) trace and determinant (phase classification) \(\sigma,\ \tau\) (swapped!)
\(e^{At}\) matrix exponential same
steady state \(\bar x\) (or \(x^\ast\)) equilibrium of \(\dot x=F(x)\) \(x^\ast\)

The course labels the trace \(\sigma\) and the determinant \(\tau\) in the planar-system classification. We use the more common \(\tau=\operatorname{tr}A\) and \(\Delta=\det A\) with discriminant \(\tau^2-4\Delta\). Watch this when comparing the \(\sigma\)\(\tau\) diagrams.

Calculus of variations (Ch. 3)

Symbol Meaning Course writes
\(J[x]=\int_0^T F(t,x,\dot x)\,\mathrm{d}t\) variational objective (functional) \(\int_0^T F\,\mathrm{d}t\)
\(F\), \(F_x\), \(F_{\dot x}\) integrand and its partials same
\(\mathcal{A}\) class of admissible paths (a subset of \(C^1[0,T]\), not of \(\mathbb{R}^n\)) \(\aleph\)
\(\delta J\), \(\delta x\) first variation, admissible variation \(\delta V\), \(\delta x\)
\(T\) terminal time (free or fixed) same
\(\rho\) continuous discount rate (\(e^{-\rho t}\)) \(r\) or \(\rho\)
\(\varphi,\psi\) terminal/boundary curves \(\varphi,\psi\)

Optimal control (Ch. 4)

Symbol Meaning Course writes
\(x\in\mathbb{R}^n\), \(u\in\mathbb{R}^m\) state, control same
\(\dot x = g(t,x,u)\) state equation (law of motion) \(\dot x=f(t,x,u)\)
\(f(t,x,u)\) running payoff \(f\)
\(H=f+\lambda g\) (present-value) Hamiltonian \(H\)
\(\lambda\) costate = shadow price of the state \(\lambda\) (a.k.a. Hamilton multiplier)
\(\mathcal{H}=f+\mu g\), \(\mu=\lambda e^{\rho t}\) current-value Hamiltonian and costate \(H_c\), \(\gamma\)
\(\eta\) multiplier on an algebraic constraint \(\theta\) (clashes with the parameter \(\theta\))

The course reuses \(\theta\) both for the multiplier on an algebraic constraint (§4.5) and for an exogenous parameter in comparative statics (§4.7). We write the multiplier as \(\eta\) and keep \(\theta\) for the parameter.

Dynamic programming (Ch. 5)

Symbol Meaning Course writes
\(V(x)\) value function \(V\)
\(\pi_t(x,u)\) per-period payoff \(\pi\)
\(x_{t+1}=f_t(x_t,u_t)\) transition same
\(\Gamma_t(x)\) feasible-action correspondence \(\Gamma\)
\(\sigma\), \(\phi\) policy; stationary Markovian policy \(\sigma\), \(\phi\)
\(\beta=\dfrac{1}{1+\rho}\in(0,1)\) discrete discount factor \(\rho\) (overloaded!)
\(T\) Bellman operator \(T\)
\(\mathbb{S}\), \(\mathbb{A}\) state space, action space same
\(\mathbb{E}_t[\cdot]\) conditional expectation given time-\(t\) info same
\(\mathrm{d}z\), Itô: \((\mathrm{d}z)^2=\mathrm{d}t\) Wiener increment same

The discount trap. The course writes the discrete-time Bellman discount and the continuous-time discount with the same symbol \(\rho\), even though one is a factor in \((0,1)\) and the other a rate in \((0,\infty)\). Throughout these notes: \(\rho\) is the continuous-time rate (\(e^{-\rho t}\), Ch. 3–4 and continuous DP) and \(\beta=1/(1+\rho)\) is the discrete-time factor (Ch. 5). Here \(\beta=1/(1+\rho)\) is the per-period (simple) factor, whereas \(e^{-\rho}\) is the continuous-compounding factor; the two agree only to first order, \(\beta=1-\rho+O(\rho^2)=e^{-\rho}+O(\rho^2)\), and the \(\rho\) in each is a distinct object (a discrete per-period rate vs. a continuous-time rate).