!EAl3      !"#$%&'()*+,-./012None"# Gsgd A dataset with elements of type a.sgdA size of the dataset.sgd_Get dataset element with a given index. The set of indices is of a {0, 1, .., size - 1} form.sgd Lazily load dataset from a disk.sgd#A dataset sample of the given size.sgdGConstruct dataset from a vector of elements and run the given handler.sgdYConstruct dataset from a list of elements, store it on a disk and run the given handler.sgdQUse disk or vector dataset representation depending on the first argument: when 3, use , otherwise use .  NoneK> sgd,Signed real value in the logarithmic domain. sgdPositive component sgdNegative component sgdSmart LogSigned constructor.sgd2Make LogSigned from a positive, log-domain number.sgd2Make LogSigned from a negative, log-domain number.sgd#Shift LogSigned to a normal domain.sgd Change the   to either negative 4 5 or positive 6 5.  None sgdGradient with nonzero values stored in a logarithmic domain. Since values equal to zero have no impact on the update phase of the SGD method, it is more efficient to not to store those components in the gradient.sgd?Add normal-domain double to the gradient at the given position.sgdDAdd log-domain, singed number to the gradient at the given position.sgd~Construct gradient from a list of (index, value) pairs. All values from the list are added at respective gradient positions.sgdConstruct gradient from a list of (index, signed, log-domain number) pairs. All values from the list are added at respective gradient positions.sgd9Collect gradient components with values in normal domain.sgd0Empty gradient, i.e. with all elements set to 0.sgdKPerform parallel unions operation on gradient list. Experimental version.7sgd!Parallel unions in the Par monad.None"#-k 8sgd3Type synonym for mutable vector with Double values.sgdVector of parameters. sgd0SGD parameters controlling the learning process."sgdSize of the batch#sgdRegularization variance$sgdNumber of iterations%sgdInitial gain parameter&sgdOAfter how many iterations over the entire dataset the gain parameter is halved'sgdDefault SGD parameter values.(sgdA stochastic gradient descent method. A notification function can be used to provide user with information about the progress of the learning.9sgd8Add up all gradients and store results in normal domain.:sgd$Scale the vector by the given value.;sgdYApply gradient to the parameters vector, that is add the first vector to the second one.(sgdSGD parameter valuessgdNotification run every updatesgdGradient for dataset elementsgdDatasetsgdStarting pointsgd SGD result !"#$%&'( !"#$%&'(None"#@<sgd3Type synonym for mutable vector with Double values.)sgdVector of parameters.*sgd0SGD parameters controlling the learning process.,sgdSize of the batch-sgdRegularization variance.sgdNumber of iterations/sgdInitial gain parameter0sgdOAfter how many iterations over the entire dataset the gain parameter is halved1sgdDefault SGD parameter values.=sgd*The gamma parameter which drives momentum.TODO: put in SgdArgs.2sgdA stochastic gradient descent method. A notification function can be used to provide user with information about the progress of the learning.>sgd+Compute the new momentum (gradient) vector.?sgd+Compute the new momentum (gradient) vector.@sgd8Add up all gradients and store results in normal domain.Asgd$Scale the vector by the given value.BsgdYApply gradient to the parameters vector, that is add the first vector to the second one.2sgdSGD parameter valuessgdNotification run every updatesgdGradient for dataset elementsgdDatasetsgdStarting pointsgd SGD result>sgdRegularization parametersgdThe parameterssgdThe current gradient?sgdThe gamma parametersgdThe previous momentumsgdThe scaled current gradient)*+,-./012 *+,-./01)2C      !"#$$%&'()*+#$$%&'()*+,-./01234/056789:7;<=89:>"sgd-0.4.0.1-3oanPNIKr6m7pSEMIYbg8eNumeric.SGD.DatasetNumeric.SGD.LogSignedNumeric.SGD.Grad Numeric.SGDNumeric.SGD.MomentumDatasetsizeelemAtloadDatasamplewithVectwithDiskwithData LogSignedposneg logSignedfromPosfromNegtoNorm toLogFloat$fNumLogSigned$fNFDataLogSigned$fOrdLogSigned $fEqLogSigned$fShowLogSignedGradaddaddLfromList fromLogListtoListempty parUnionsParaSgdArgs batchSizeregVariterNumgain0tausgdArgsDefaultsgdghc-prim GHC.TypesTruebase Data.EitherLeft(logfloat-0.13.3.3-9PEBcHsUZGmG51I28knud7Data.Number.LogFloatLogFloatRight parUnionsPMVectaddUpscaleaddTogammaapplyRegularizationupdateMomentum