API
Theseus.combineSorted — MethodReturns a sorted matrix of points based on their indices in the original matrix. This allows for changes to some elements of the matrix in a differentiable way since zygote does not support mutating operations and the typical assignment operator mutates.
Theseus.crossEntropyLoss — MethodWeighted cross entropy loss, best used with a plain mask vector of weights.
Theseus.crossEntropyLoss — MethodCross entropy loss function.
Theseus.forceVar — MethodReduce the difference between the maximum and minimum forces in the network. From Schek theorem 2.
Theseus.lenVar — MethodCompute difference between the maximum and minimum lengths of the edges in the network.
Theseus.logSumExp — MethodLogSumExp is the multivariable generalization of the logistic function.
Theseus.logisticFunc — MethodThe derivative of the softplus function is the logistic function. A scaling parameter k can be introduced to make the inflection point more precise. This is a smooth approximation of the heaviside step function. https://en.wikipedia.org/wiki/Logistic_function
Theseus.maxPenalty — MethodPenalizes values in vector that are above a threshold
Theseus.minPenalty — MethodPenalizes values in vector that are below a threshold
Theseus.pairDist — MethodFind the distance between all pairs of points in a point set. Returns a strictly lower triangular matrix.
Theseus.rigidSetCompare — MethodCompare the distance between all pairs of points in a target point set and the distance between all pairs of points in a form found point set.
Theseus.softmax — MethodSoftmax is a generalized version of the logistic function. It returns a vector of probabilities that sum to 1.
Theseus.softplus — MethodSoftplus is a smooth approximation of the ReLU function. A sharpness parameter k may be included to make the inflection point more precise.
x is the input, b is the inflection point bias, k is the sharpness parameter for the barrier slope.
negative k raises a barrier on the left side of the inflection point. positive k raises a barrier on the right side of the inflection point.
Theseus.target_xy — MethodMinimize the distance between the x and y coordinates of the target nodes and their corresponding nodes in the form found network. Equal to targeting a plan projection of the target nodes. Useful if the target geometry variation is dominated by the x and y coordinates.
Theseus.target_xyz — MethodMinimze distances between selected target nodes and their corresponding nodes in the form found network.