I'm trying to use the Graphical Lasso algorithm (more specifically the R package glasso) to find an estimated graph representing the connections between a set of nodes by estimating a precision matrix. I have a feature matrix containing the values of multiple features for each of the nodes, and the sample covariance matrix obtained from the product between this matrix and its tranpose is used as the input for the glasso function, along with the l1 regularization coefficient $\lambda$. However, …

Is there any tool to visualize the feasible region when given a set of Linear equations (equalities and inequalities). If not, can anyone suggest a way to visualize it? If I am going to do it myself using Python, which libraries should I use. I have found sympy, but I couldn't get it to draw inequalities nor draw the intersections only. I have also found wolfram, but I could only see pre-built visualizations and not visualize my own system. Can …

I wish to obtain the most probable set of DAGs created from observational data and then compare them using my own score function. Ideally, I would want to give the observational data and then the number of candidate DAGs I would want in return (n) as input and get the 'most appropriate' n DAGs. Is this possible?

Boltzmann machines were introduced by Hinton and Sejnowski as taking values in $\{0,1\}$. The Wikipedia entry also uses this convention. However, Hopfield Networks, which are the deterministic version of Boltzmann machines, are usually introduced as taking values in $\{-1,1\}$. Ising models also follow this convention. With the energy function being defined equivalently in both models as $$ E(x) = \sum_i b_ix_i + \sum_{i<j}w_{ij}x_ix_j$$ it seems that the two conventions would behave differently. For example, how would the $\{0,1\}$ model learn …

Pedro Domingos in "The Master Algorithm" listed five tribes of machine learning algorithms: Symbolists Connectionists Evolutionaries Bayesians Analogizers Which category do probabilistic graphical models fall under? From wikipedia (https://en.wikipedia.org/wiki/Graphical_model): A graphical model or probabilistic graphical model (PGM) is a probabilistic model for which a graph expresses the conditional dependence structure between random variables. They are commonly used in probability theory, statistics—particularly Bayesian statistics—and machine learning. In that case, would it be "Bayesian"? What about something like Conditional Random Field? Is …

I would have thought that it's because DAGs preserve the dependency relationships between the variables, but I am currently unsure. Thanks

On Bayesian Networks, Ghahramani (2001) says: A node is independent of its non-descendants given its parents. This point is fundamental enough that Ghahramani calls it the “semantics” of a Bayesian network. It is certainly useful, and it is simple enough to prove using d-separation. But his characterization suggests that the property should be even more primitive than something provable by d-separation. Overall, I feel that I am missing something. Is there a more primitive way to verify the statement than …

I need to establish if there is a link between 2 columns from two different datasets with one matching column, where; Dataset1: bipartite: (M, DS) M G m23 ds3 m23 ds67 m54 ds325 ... ... Dataset2: tripartite: (M, G, DG) M G DG m23 g6 dg32 m23 g8 dg1 m54 g32 dg65 ... ... ... These 2 datasets have one column in common(i.e., M), and the relationship among the elements is shown below: M ----affects----> G M ----causes-----> DS DG …

I want to create a dependency graph of some sensors in the network ( based on their reported value). Please note that a change in the values of sensors is related to each other. For example, if the measured value by a sensor increases, it is possible to recognize the sensors which their measured values will be affected by this increase. This is because of the mathematical logic behind the reported values and sensor placement. I also need to have …

I'm following this tutorial here to build an ARIMA model in R. I've done a Forecast using a fitted model in R. I specified the forecast horizon h periods ahead for predictions to be made and used the fitted model to generate those predictions. Then I plotted them to see the results and this is what I got: The light blue line above is supposed to show the fit provided by the model. In the Tutorial this is what they …

Is there any good libraries that allow me to: Construct a Bayesian network manually Specify the conditional probabilities with any continuous PDF, not just Guassian Perform inference, either exact or approximate I looked at the following libraries so far, none of them meet the 3 requirements: pgmpy: only work on discrete distribution or linear Guassian distribution bnlearn: same as pgmpy gRain: only discrete distribution Huggin: only discrete distribution and Guassian deal: no support for inference abn: same as deal libpgm: …

I have a dataset that contains the population of butterflies(5 species) for 15 years for different locations. I want to model it against the climate index collected for same time period and location. The objective is to find how early the migration species migration starts before the onset of a dry period? My idea was to understand all the spatial locations around the dry place and see if the populations have changed over time in the neighbourhood which can indicate …

I built my first decision tree, to predict if students will pass or not - the data set - depending on 30 variables. Now I need to know how to read the decision tree, since many variables were strings and now are floats made me confused. The 30 variables are: school sex age address famsize Pstatus Medu Fedu Mjob Fjob reason guardian traveltime studytime failures schoolsup famsup paid activities nursery higher internet romantic famrel freetime goout Dalc Walc health absences …

I am trying to find an interesting way to interpret and display a set of data for the research I'm working on. Columns 2-4 show the net change from time 1 to time 2 in antibiotic coverage for different types of bacteria (types A-D). 1 means coverage for that type of bacteria was added, and -1 means coverage for that bacteria was stopped, and 0 means there was no change in coverage (i.e. it wasn't covered at time 1 or …

Traditional Viterbi algorithm (say, for hidden Markov models) provides the most probable hidden state sequence given a sequence of observations. There probably is an algorithm for decoding top-N probable hidden states sequences (k-shortest paths or the like). But is there a good implementation anywhere? Thanks. UPD (copy-pasted from comments): Viterbi algorithm decodes the MOST probable sequence given observations + model. That is, argmax_x p(state_1,...,state_n|obs_1,...,obs_n). What I am asking for is an implementation of how to get the 'next argmax-es' with …

I currently have a Graph model whereby I am mapping connections of different types between entities and attributing a weight to these connections based upon my own personal experience. Also, I would like to understand the connections between these entities in relation to a particular outcome. Looking at this problem, I can't help but notice it's similarity to a typical Artificial Neural Network (ANN) and am wondering if/how I can bake some of the theory there into my model. Let …

Can anyone explain (or refer to a great explanation of) the intuition of how Latent SVM works? I think Latent SVM should have some resemblance to CRF (Conditional Random Fields) and EM (expectation maximization) and of course standard SVM (max-margins, kernel trick) and I'll appreciate very much answers using these 'tools' to explain how Latent SVM works.

I basically want to build a system which will provide a student step by step guide or you can say a full route about what courses he/she should take currently or what examinations they need to clear to ultimately land up in universities of their choice in future. I want to this project for the process of learning and i need some starting points about how should i go about doing this .

I have the following problem: Suppose we are interested in estimating the distribution over the English letters. We assume an alphabet that consists of 26 letters and the space symbol, and we ignore all other punctuation and the upper/lower case disctinction. We model the distribution over the 27 symbols as a multinomial parametrized by $\theta =(\theta_1,...,\theta_{27})$ where $\sum_i \theta_i = 1$ and all $\theta_i \geq 0$/ Now we go to Stanford's Green Library and repeat the following experiment: randomly pick …

Here is a simple example: In a 3D space, if point A is the geocenter of a planet, point B is its north pole, and point C has a fixed latitude/longitude on the planet surface. Then the position of point C can be inferred from the position of A and B (using geodetic projection). Now assuming that I have thousands of points in the space, their incomplete relations are in a big multipgraph. I would like to deduce the position …