blarney/rubble

Tony Coulter here, with a second post from my new digs in Portland, Oregon. I should say right off the bat that if I write about Portland, my intent is not to promote the place as the perfect spot for a hipster jamboree. I happen to have just moved here, so, naturally, Portland's on my mind; if I had moved to Sandpoint, Idaho, I'd be writing about Sandpoint, Idaho. In fact, to prove it, I'll just go ahead and throw in two things Idahoan. After all, Idaho and Oregon are neighbors, aren't they?

Anyhow, I was originally planning to do a post on some of Oregon's musical highlights (from my perspective, of course). I abandoned that as too broad a topic, but if I had done it, I'd have urged you to listen to the Lollipop Shoppe, Beauregarde, the New Tweedy Bros., the New Dawn, Hunger, Smegma, Jungle Nausea, the Parasites of the Western World, and Jackie-O Motherfucker, among others. Digging deeper, I would have recommended Delvin Ford, Bob Desper, and Nun-Plus. 

All of the above have actually already been played on FMU -- so you can check them out by trolling archives of yore.  What I decided to do instead was to highlight some LPs from Oregon and Idaho that I've discovered since moving here four months ago. These are things I would have played on my FMU show if I were still on the air. I will also show you pictures from a high school yearbook I found in a junk shop, and from a book put out by a Berkeley dentist.

Sounds and sights await you past the fold:

  First up is a fine unknown Portland LP, released in 1981 and camouflaged behind a very drab cover. Overseen by group leader Richard Crandall -- rhythm guitarist, keyboardist, vocalist, and inventor of a digital piano called the "Albatross" -- the Chameleons cover a pretty broad range of styles, from pseudo-country to post-punk. There's an underlying artiness and eggheadedness, and unusually interesting lyrics. I've chosen two of the post-punkier tracks (even if the lead guitarist hadn't gotten the memo), which you'll find below the portrait of Herr Crandall and his digital sidekick.

Richard Crandall with the "Albatross"

The Chameleons: Men Are Like Snowflakes

The Chameleons: Escape

__________________________________________________________________________________________

Stauz: My Stars! (Stauf Enterprises, 1981)

  Moving on to a second disc from Portland -- and also from 1981 -- we enter trickier aesthetic waters. The duo known as Stauz has definite outsider, "real people" qualities. That is to say, they don't possess much in the way of conventional musical skill, they are frequently "out of tune," and their songs have a strange, somewhat unhinged obsessive quality. Nonetheless, they are unique and I genuinely like them -- I hope you will too. As Stauz themselves remark, "Jealousy has absolutely nothing whatsoever to do with the science or philosophy of jello, jellocy."

 

 

Stauz: A Jealous Rage of Blue

Stauz: Prayers to God

___________________________________________________________________________________________

The Mission Street Salvation Band: Adriel (Flock of God) (New World NW 1004, 1973)

 Next up is a 1973 Jesus rock LP by a band from Dallas, Oregon (a town not too far from Portland, but closer to Salem). This is heartfelt, like the best of this kind of stuff, and has a nice laid back quality. The track I've chosen has a rather stoned-out groove and features striking harmonica playing, which is unusual for not really being blues-based. The atmospheric shot of the band on the back cover both reminds me of the cover of the New Dawn LP, and suggests the band probably had real roots in the counterculture -- that is, they probably found Jesus on an acid trip.

 

 The Mission Street Salvation Band: Door at the Foot of the Cross

__________________________________________________________________________________________

Jim Ballou: Key to a Fantasy (Musicmaker Productions, 1988)

  Crossing the border into Idaho, we turn next to a fascinating record from Sandpoint, self-released in 1988. Like Stauz, Ballou is perhaps an "outsider" artist -- but, probably unintentionally, there's also a kind of echo here of certain kinds of faux-guileless indie rock. Ballou, however, is completely sincere, and the album has a delicate and touching charm.

 Jim Ballou: Only in a Daydream

Jim Ballou: She'll Remember Me

__________________________________________________________________________________________

Stone Garden: Stone Garden (Gear Fab GF-188, 2002)

  Last month, while rooting around a local Portland junk shop, I came across a 1970 yearbook from a high school in Lewiston, Idaho. Flipping through it, I was startled to discover a picture (see below) of just about the only psych band from Idaho I'm familiar with: The Stone Garden. Apparently, they played the graduation dance! Perhaps they even played the song about the end of the world I've posted underneath their action shot. 

 

Stone Garden: The World Is Coming to an End

__________________________________________________________________________________________

Lewiston High School, Lewiston, Idaho - The 1970 Bengal

The "Stone Garden" yearbook has other charms as well. It's quite fascinating flipping through it to see how much the '50s and '60s overlapped in that time and place. Loafers and pom-poms rub shoulders with shaggy hair and swirly psychedelic art. Talking about swirly psychedelic art, here's my favorite example, taken from the yearbook's flyleaf and drawn by someone whose name I can't quite make out:

Artwork: Karen P.

_________________________________________________________________________________________

Thomas McGuire D.D.S.: The Tooth Trip  (Random House/Bookworks, 1972)

  Turning next to a book I found awhile ago, it occurs to me that I'm not sure I'd want an acid head for a dentist -- but in Berkeley in 1972, pretending to be an acid head apparently struck Doctor Thomas McGuire as a good way to attract customers. How else to explain the fisheye lens and the Blue Cheer album cover on the ceiling?  In case you've forgotten what Blue Cheer's Outsideinside looks like, I've included a photo of a copy "Lynette" bought for ten cents.

Here's the book's title page:

Lynette's ten-cent copy of Blue Cheer's Outsideinside

 _________________________________________________________________________________________

That's it for this week's installment. In two weeks time: an interview with Smegma!

Having been motivated, today we are going to walk through a small example and touch on the main concepts related to learning within this library.

At the top of food chain is the generic function TRAIN:

 (defgeneric train (sampler trainer learner)
   (:documentation "Train LEARNER with TRAINER on the examples from
 SAMPLER. Before that TRAINER is initialized for LEARNER with
 INITIALIZE-TRAINER. Training continues until SAMPLER is finished."))

A learner is anything that can be taught, which currently means it's either a backpropagation network (BPN) or some kind of boltzmann machine (BM). The method with which a learner is trained is decoupled from the learner itself and lives in the trainer object. This makes it cleaner to support multiple learning methods for the same learner: for instance, either gradient descent (BP-TRAINER) or conjugate gradients (CG-BP-TRAINER) can be used to train a BPN, and either contrastive divergence (RBM-CD-TRAINER) or persistent contrastive divergence (BM-PCD-TRAINER) can be used to train a restricted boltzmann machine (RBM).

The function TRAIN takes training examples from SAMPLER (observing the batch size of the trainer, if applicable) and calls TRAIN-BATCH with the list of examples, the trainer and the learner. This may be as simple as:

 (defmethod train (sampler (trainer bp-trainer) (bpn bpn))
   (while (not (finishedp sampler))
     (train-batch (sample-batch sampler (n-inputs-until-update trainer))
                  trainer bpn)))

Ultimately, TRAIN-BATCH arranges for the training examples to be given as input to the learner ("clamped" on the input nodes of some network) by SET-INPUT; exactly how this should be done must be customized. Then, in the case of BP-TRAINER, the gradients are calculated and added to the gradient accumulators that live in the trainer. When the whole batch is processed the weights of the network are updated according to the gradients.

Let's put together a toy example:

 (use-package :mgl-util)
 (use-package :mgl-train)
 (use-package :mgl-gd)
 (use-package :mgl-bp)
 
 (defclass linear-bpn (bpn) ())
 
 (defparameter *matrix*
   (matlisp:make-real-matrix '((1d0 2d0) (3d0 4d0) (5d0 6d0))))
 
 (defparameter *bpn*
   (let ((n-inputs 3)
         (n-outputs 2))
     (build-bpn (:class 'linear-bpn)
       (input (input-lump :size n-inputs))
       (weights (weight-lump :size (* n-inputs n-outputs)))
       (product (activation-lump :weights weights :x input))
       (target (input-lump :size n-outputs))
       (sse (->sum-squared-error :x target :y product))
       (my-error (error-node :x sse)))))
 
 (defmethod set-input (samples (bpn linear-bpn))
   (let* ((input-nodes (nodes (find-lump 'input bpn)))
          (target-nodes (nodes (find-lump 'target bpn)))
          (i-v (storage input-nodes)))
     (assert (= 1 (length samples)))
     (loop for i below (length i-v) do
           (setf (aref i-v i) (elt (first samples) i)))
     ;; TARGET-NODES = INPUT-NODES * *MATRIX*
     (matlisp:gemm! 1d0 (reshape2 input-nodes 1 3) *matrix*
                    0d0 (reshape2 target-nodes 1 2))))
 
 (defun sample-input ()
   (loop repeat 3 collect (random 1d0)))
 
 (train (make-instance 'counting-function-sampler
                       :sampler #'sample-input
                       :max-n-samples 10000)
        (make-instance 'bp-trainer
                       :segmenter
                       (repeatedly
                         (make-instance 'batch-gd-trainer
                                        :learning-rate (flt 0.01)
                                        :momentum (flt 0.9)
                                        :batch-size 10)))
        *bpn*)

We subclassed BPN as LINEAR-BPN and hanged a SET-INPUT method on it. The SAMPLES argument will be a sequence of samples returned by the sampler passed to TRAIN, that is, what SAMPLE-INPUT returns.

The network multiplies INPUT taken as a 1x3 matrix by WEIGHTS (initialized randomly) and the training aims to minimize the squared error as calculated by the lump named SSE. Note that SET-INPUT clamps both the real input and the target.

We instantiate BP-TRAINER that inherits from SEGMENTED-GD-TRAINER. Now, SEGMENTED-GD-TRAINER itself does precious little: it only delegates training to child trainers where each child is supposed to be a GD-TRAINER (with all the usual knobs such as learning rate, momentum, weight decay, batch size, etc). The mapping from segments (bpn lumps here) of the learner to gd trainers is provided by the function in the :SEGMENTER argument. By using REPEATEDLY, for now, we simply create a distinct child trainer for each weight lump as it makes a function that on each call evaluates the form in its body (as opposed to CONSTANTLY).

That's it without any bells and whistles. If all goes well WEIGHTS should be trained to be equal to *MATRIX*. Inspect (nodes (find-lump 'weights *bpn*)) to verify.

Impatience satisfied, examine the BUILD-BPN form in detail. The :CLASS argument is obvious, and the rest of the forms are a sequence of bindings like in a LET*. The extra touches are that the name of the variable to which a lump is bound is going to be supplied as the :NAME of the lump and an extra MAKE-INSTANCE is added so

 (input (input-lump :size n-inputs))

is something like

 (make-instance 'input-lump :name 'input :size n-inputs)

One can replicate this with MAKE-INSTANCE and ADD-LUMP, but it's more work. For ease of comprehension the network can be visualized by loading the mgl-visuals system and:

 (let ((dgraph (cl-dot:generate-graph-from-roots *bpn* (lumps *bpn*))))
   (cl-dot:dot-graph dgraph "linear-bpn.png" :format :png))

That's it for today, thank you for your kind attention.

Visit msnbc.com for Breaking News, World News, and News about the Economy