numpy.abs

Eigenvalues with numpy.linalg: normal modes of coupled oscillators

• residual = np.abs(K / m @ V - V @ np.diag(w2)).max()
• print(f"largest entry of V^T V - 1: {np.abs(V.T @ V - np.eye(3)).max():.0e}")
• v = v / np.abs(v).max()

Filtering with scipy.signal: mains hum and noise out of an ECG

• return 2 * np.abs(np.fft.rfft(s)) / s.size # mV
• k50 = np.argmin(np.abs(freq - 50))
• print(f"harmonic {n:2d} {A_raw[np.argmin(np.abs(freq - f_n))]:.4f} mV at {f_n:.1f} Hz")
• for f_c, g in zip(f_check, 20 * np.log10(np.abs(h))):
• gain = 20 * np.log10(np.abs(h))
• print(f"{N:5d}" + "".join(f"{g:8.1f} dB" for g in 20 * np.log10(np.abs(h))))

and 8 more lines in this tutorial

Minimization with scipy.optimize.minimize: the shape of a seven-atom cluster

• print(f"largest difference to approx_fprime: {np.abs(g - approx_fprime(x5, energy)).max() / np.abs(g).max():.1e} of the largest component")
• label = f"{energies[np.abs(energies - e) < 0.03].mean():.3f}".replace("-", "\u2212") # minus sign as on the axis

Numerical integration with scipy.integrate: the area under a measured peak

• win = np.abs(lam - peak) <= half + 1e-9 # the slack keeps the edge points despite rounding
• f = lambda x: 1 / np.sqrt(np.abs(x))

Random numbers with numpy.random: ten thousand reproducible random walks

• outside = n[np.abs(msd_n - n) > band]

py-pde from the ground up: the heat equation on a square plate

• deviation[t] = np.abs(field.data - exact[:, None]).max()
• print(f"t = 20 max |T - (1 - x)| = {np.abs(T_flux.data - exact[:, None]).max():.3g}")
• dev_flux = {t: np.abs(f.data - exact[:, None]).max() for t, f in storage_flux.items()}
• print(f"largest |T| after 100 steps: {np.abs(T128.data).max():.1e}")

solve_ivp from the ground up: the pendulum beyond small angles

• drift = np.max(np.abs(E - E[0])) / E[0]