jparkhill

def kelly_loss(c_assets, r_assets, spreads, spread_loss):
    """
    Maximizes exponential rate of wealth growth 
    
    NOTE: the samples must be in the same order. 
          the assumption is that the returns are realized 
          in the same way. 
    
    Args: 

jparkhill

    // Do a shifted/scaled fourier transform of polarization in each direction.
    mat Fourier(const mat& pol, double zoom = 1.0/20.0, int axis=0, bool dering=true) const
    {
        cout << "Computing Fourier Transform ..." << endl;
        mat tore;
        double dt = pol(3,0)-pol(2,0);
        {
            int ndata = (pol.n_rows%2==0)? pol.n_rows : pol.n_rows-1; // Keep an even amount of data
            for (int r=1; r<ndata; ++r)

jparkhill

def slerp(v0, v1, t=0.05):
    """
    Interpolate between quaternions v0 and v1.
    """
    v0 = safe_normalize(v0)
    v1 = safe_normalize(v1)
    dot = tf.reduce_sum(v0*v1,axis=-1,keepdims=True)
    # If the dot product is negative, slerp won't take
    # the shorter path. Note that v1 and -v1 are equivalent when
    # the negation is applied to all four components. Fix by

jparkhill

// ****************************************************************************
// Customizable Smartphone Holders
// to mount your smartphone to a tripod or elsewhere using an action cam hook
// Author: Peter Holzwarth
// ****************************************************************************
// https://en.wikibooks.org/wiki/OpenSCAD_User_Manual/The_OpenSCAD_Language

SmartphoneHolder([PhoneName, PhoneWidth, PhoneHeight, PhoneDepth,
    PhoneCornerRoundingRadius, PhoneEdgeRoundingRadius, 
    PhoneCamPos, CamSlotHeight], MountPos);

jparkhill

class Neural_Density(torch.nn.Module):
    """
    A neural model of a time dependent probability density on 
    a vector valued state-space.
    
    ie: rho(t,{x_0, x_1, ... x_{state_dim}})
    
    for now, I'm not even enforcing normalization. 
    could with a gaussian mixture or whatever. 
    """