Skip to content

anypinn.cli.scaffold.lorenz.ode_csv

Lorenz system — mathematical definition.

BETA_KEY = 'beta' module-attribute

NOISE_STD = 0.5 module-attribute

RHO_KEY = 'rho' module-attribute

SCALE = 20.0 module-attribute

SIGMA_KEY = 'sigma' module-attribute

TRUE_BETA = 8.0 / 3.0 module-attribute

TRUE_RHO = 28.0 module-attribute

TRUE_SIGMA = 10.0 module-attribute

T_TOTAL = 3 module-attribute

X0 = -8.0 module-attribute

X_KEY = 'x' module-attribute

Y0 = 7.0 module-attribute

Y_KEY = 'y' module-attribute

Z0 = 27.0 module-attribute

Z_KEY = 'z' module-attribute

validation: ValidationRegistry = {} module-attribute

create_data_module(hp: ODEHyperparameters)

Source code in src/anypinn/cli/scaffold/lorenz/ode_csv.py 
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
def create_data_module(hp: ODEHyperparameters):
    from anypinn.catalog.lorenz import LorenzDataModule

    def lorenz_unscaled(x: Tensor, y: Tensor, args: ArgsRegistry) -> Tensor:
        lx, ly, lz = y
        sigma = args[SIGMA_KEY]
        rho = args[RHO_KEY]
        beta = args[BETA_KEY]
        dx = sigma(x) * (ly - lx)
        dy = lx * (rho(x) - lz) - ly
        dz = lx * ly - beta(x) * lz
        return torch.stack([dx, dy, dz])

    gen_props = ODEProperties(
        ode=lorenz_unscaled,
        y0=torch.tensor([X0, Y0, Z0]),
        args={
            SIGMA_KEY: Argument(TRUE_SIGMA),
            RHO_KEY: Argument(TRUE_RHO),
            BETA_KEY: Argument(TRUE_BETA),
        },
    )

    return LorenzDataModule(
        hp=hp,
        gen_props=gen_props,
        noise_std=NOISE_STD,
        validation=validation,
        callbacks=[DataScaling(y_scale=[1 / SCALE, 1 / SCALE, 1 / SCALE])],
    )

create_problem(hp: ODEHyperparameters) -> ODEInverseProblem

Source code in src/anypinn/cli/scaffold/lorenz/ode_csv.py 
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
def create_problem(hp: ODEHyperparameters) -> ODEInverseProblem:
    props = ODEProperties(
        ode=lorenz_scaled,
        y0=torch.tensor([X0, Y0, Z0]) / SCALE,
        args={},
    )

    fields = FieldsRegistry(
        {
            X_KEY: Field(config=hp.fields_config),
            Y_KEY: Field(config=hp.fields_config),
            Z_KEY: Field(config=hp.fields_config),
        }
    )
    params = ParamsRegistry(
        {
            SIGMA_KEY: Parameter(config=hp.params_config),
            RHO_KEY: Parameter(config=hp.params_config),
            BETA_KEY: Parameter(config=hp.params_config),
        }
    )

    def predict_data(x_data: Tensor, fields: FieldsRegistry, _params: ParamsRegistry) -> Tensor:
        x_pred = fields[X_KEY](x_data)
        y_pred = fields[Y_KEY](x_data)
        z_pred = fields[Z_KEY](x_data)
        return torch.stack([x_pred, y_pred, z_pred], dim=1)

    return ODEInverseProblem(
        props=props,
        hp=hp,
        fields=fields,
        params=params,
        predict_data=predict_data,
    )

lorenz_scaled(x: Tensor, y: Tensor, args: ArgsRegistry) -> Tensor

Scaled Lorenz ODE. States pre-divided by SCALE, time by T_TOTAL.

Source code in src/anypinn/cli/scaffold/lorenz/ode_csv.py 
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
def lorenz_scaled(x: Tensor, y: Tensor, args: ArgsRegistry) -> Tensor:
    """Scaled Lorenz ODE. States pre-divided by SCALE, time by T_TOTAL."""
    lx, ly, lz = y
    sigma = args[SIGMA_KEY]
    rho = args[RHO_KEY]
    beta = args[BETA_KEY]

    dx = sigma(x) * (ly - lx)
    dy = lx * (rho(x) / SCALE - lz) - ly
    dz = lx * ly * SCALE - beta(x) * lz

    dx = dx * T_TOTAL
    dy = dy * T_TOTAL
    dz = dz * T_TOTAL
    return torch.stack([dx, dy, dz])