evoldoers / machineboss Goto Github PK

this will make it easier to do 1D^2 (intersection of generator w/two parallel machines)

Only those into cpt_end states. (And specifically not those into end.)
This is causing discrepancies between Viterbi matrix score & path score.
Need to explicitly flag some states as softplus-able.

and/or skip estimation of TraceParams

Use Poreover example data

• Change TransWeight -> WeightFunc
• typedef map<string,json> FuncDefs
• Change Params in WeightAlgebra to FuncDefs
• use change of variables instead of Lagrange multipliers

• Envelope data structure
• Initialize from MachinePath
• connected method (tests all start/end intervals overlap & touch corners)
• Adapt containers to use packed storage
• Adapt iterators (this is really #18)
• SeqPair path property (can be empty)

Banded initializer is now #66

This means porting EvaluatedMachine, DPMatrix, ViterbiMatrix

As #33, but using the compiler.

function (subclass/compatible with #8) capable of

• eval
• deriv
• constraint (Lagrange multiplier)

• Implement log_sum_exp directly in generated code
• Convert probability profiles to sequences on-the-fly

Needs #14, #17

Input and/or output can be a sequence instead of a profile

e.g. for protein to RNN, input is a sequence (protein) and output is a profile

All cases (conditioned on input, output, both, or neither) can be reduced to the null machine (empty input & output) by appropriate compositions with generators/acceptors.
Then, that null machine can be solved by matrix inversion. In fact, the machine is acyclic & topologically sorted at this point, so the inversion can be done efficiently by calculating the sum-over-paths likelihood from each state to the end state.

Subclass of MachineCounts initialized from ForwardMatrix + BackwardMatrix. Needs #13

Serializing final model to JSON is rate-limiting step.

Compiled & uncompiled. To test gap handling.

To reproduce:

cd src/nano/test
biomake -B seq1-test1.decode |& less -r

Like #43, but compiled.

e.g.
nanomachine -R seq1-test1.raw -M test1.json -v8 -C -d TraceViterbiMatrix -f .00001
should give same results as
nanomachine -R seq1-test1.raw -M test1.json -v8 -C -d TraceViterbiMatrix -g 1
but doesn't

e.g. RNN outputs

Traceback yields alignments - needs #14

• container for MachineTransitionScore
• requires #10

As it currently does; but keep the names

As with #34, Machine-creating operations should propagate/merge their ParamDefs

• Uses GSL maximizer with constrained parameter prior
• requires #11 #12 #13 #15

• feasibility calculations (CPU, memory)
• training
• search

With evaluation (when loaded)
Special Machine constructor

logspace

It should be proportional to the number of backward silent transitions that have been eliminated, not the number of states visited

All unary & binary Machine operations should propagate/merge Constraints

equivalent to calculating sequence probability using CTC (Graves et al)

• feasibility calculations (CPU, memory)
• test data (c.f. #32)

Start with bitnoise test

Interpreted, compiled

• logspace_eval = log(f(x1,x2...))
• logspace_deriv = dlog(f(x1,x2...))/dlog(xn)