markov
markov v4.1.3

Markov.ModelActions (markov v4.1.3)

Performs training, generation and probability shifting. Supposed to only ever be used by Markov.ModelServers.

Link to this section Summary

Functions

adjust_batch_probs(params)

Adjust the probabilities of a batch of connections

adjust_one_prob(param_tensor)

Adjusts the probability of one connection

generate(state, tag_query)
next_state(state, current, tag_query)
process_scores(rows, tag_scores)

processes tag scores

train(state, tokens, tags)
walk_chain(state, acc, queue, limit, tag_query)

Link to this section Functions

Link to this function

adjust_batch_probs(params)

Adjust the probabilities of a batch of connections

Link to this function

adjust_one_prob(param_tensor)

Adjusts the probability of one connection

Link to this function

generate(state, tag_query)

@spec generate(Markov.ModelServer.State.t(), Markov.tag_query()) ::
  {{:ok, [term()]} | {:error, term()}, Markov.ModelServer.State.t()}
Link to this function

next_state(state, current, tag_query)

@spec next_state(Markov.ModelServer.State.t(), [term()], Markov.tag_query()) ::
  {:ok, term(), Markov.ModelServer.State.t()}
  | {:error, term(), Markov.ModelServer.State.t()}
Link to this function

process_scores(rows, tag_scores)

@spec process_scores([{term(), non_neg_integer(), term()}], Markov.tag_query()) :: %{
  required(term()) => non_neg_integer()
}

processes tag scores

Link to this function

train(state, tokens, tags)

@spec train(
  state :: Markov.ModelServer.State.t(),
  tokens :: [term()],
  tags :: [term()]
) :: :ok
Link to this function

walk_chain(state, acc, queue, limit, tag_query)

@spec walk_chain(
  Markov.ModelServer.State.t(),
  [term()],
  [term()],
  non_neg_integer(),
  Markov.tag_query()
) :: {{:ok, [term()]} | {:error, term()}, Markov.ModelServer.State.t()}