Independent scholarship · Open accessThe monthly digest

Indexed text attachment

differential-topology-of-parametrized-neural-networks.tex

Source for Differential Topology of Parametrized Neural Networks

View paper · Download original file

Lines 254–353 of 1074. Source text is displayed literally.

254projection to $M$. For $J^0$ only the value variation is255needed, and free biases supply it at every $x$.256 257A submersion is transverse to each smooth jet submanifold $S$.258Its preimage is smooth, and regular values of the projection259of the preimage to $\mathcal A$ are exactly the parameters260whose jet extensions are transverse to $S$. This is the261parametric transversality argument262\cite[Ch.~6]{higher-hirsch}. Sard's theorem gives a null263exceptional set. The critical locus has a countable compact264exhaustion because it is closed in a second-countable265manifold. Its images under the projection are compact266null sets and therefore nowhere dense. Taking the countable267intersection of the good parameter sets proves the claim.268\end{proof}269 270For a binary classifier, Theorem~\ref{thm:jet-surjectivity}271implies that generically the decision boundary is empty or a272codimension-one smooth submanifold, even if $\phi$ has273singular points. The free final bias alone suffices. A274fixed-bias family can have additional constraints.275For example, if a feature identically vanishes on a276region, a bias-free readout may vanish on that entire region.277For vector outputs, immersion of $\phi$ supplies every278first-order normal perturbation at a given input. It makes279no assertion that independent perturbations at multiple280inputs can be realized by the same $\dot A$.281 282\subsection{Several classes and several points}283 284Consider $q$ logits and the open part of the tie stratum285where a subset $I$ of $r$ classes is exactly the set of286maximizers. Fix $i_0\in I$. Equalities287$y_i-y_{i_0}=0$ for $i\in I\setminus\{i_0\}$ are $r-1$288independent linear equations, and strict inequalities289separate the other coordinates. This is a smooth290codimension-$(r-1)$ submanifold of $\mathbb R^q$.291Consequently, the inverse image for a generic readout292is empty or has codimension $r-1$. No $r$-way maximum293tie occurs if $r>n+1$. The conclusion also holds when294the bias space is restricted to vectors with coordinate295sum zero. The differences of any bias vector are296unchanged by subtracting a constant vector so the297variation still surjects onto the normal directions to298each tie. Thus the ordinary redundancy of adding the299same constant to every logit can be removed without300losing the relevant perturbations.301 302There is a complementary global result when $\phi$303is an embedding. For different points $x,x'$ the304map $(A,x,x')\mapsto A(\phi(x)-\phi(x'))$ is a submersion305to $\mathbb R^q$. Indeed, $\phi(x)-\phi(x')$ is nonzero306and a linear map can assign it any value. The source307in the input variables is $M\times M\setminus\Delta$,308of dimension $2n$. When $q>2n$, parametric309transversality forces absence of off-diagonal coincidences310for a conull residual set of $A$. By the first-jet argument,311$r$-rank derivative strata of codimension312$(n-r)(q-r)$ are avoided when that codimension exceeds $n$.313This formula follows by choosing an invertible $r$-minor314and using block elimination: the remaining315$(q-r)\times(n-r)$ block after elimination must be zero.316In particular for $q\ge2n$ all nonimmersive strata317are avoided. For compact $M$ and $q\ge2n+1$ the generic318readout of an embedding is thus an embedding. An affine319bias has no effect on this two-point test. This is a version320of the projection argument for embeddings applied to321the allowed neural readouts.322 323\subsection{A two-parameter boundary discriminant}324\label{sec:threshold-discriminant}325 326Use the unit circle as a data manifold with coordinate327$x=(\cos t,\sin t)$. A network with one square unit and328affine readout can realize329\[330 F(a,b,t)=\cos t+a\cos(2t)-b331\]332because $\cos(2t)=2(\cos t)^2-1$.333Its zero locus in $(a,b,t)$ is smooth since the last bias334derivative is $-1$. Projection to $(a,b)$ is proper335because the circle is compact. A parameter belongs336to the discriminant when a root also satisfies337\[338 \partial_t F=-\sin t(1+4a\cos t)=0.339\]340Thus the discriminant consists of the two lines341$b=a+1$ and $b=a-1$ and the curve342$b=-a-1/(8a)$ for $|a|\ge1/4$.343At $|a|=1/4$ the latter joins a line with a degenerate344critical point. At $a=1/4$ and $t=\pi$ the first nonzero345variation of $\cos t+a\cos 2t$ in $t-\pi$ is quartic,346with coefficient $1/8$.347 348\begin{figure}[tbp]349\centering350\begin{tikzpicture}[x=2.3cm,y=.86cm,font=\small]351 \draw[->,gray!85] (-1.04,0) -- (1.08,0) node[right] {$a$};352 \draw[->,gray!85] (0,-2.02) -- (0,2.03) node[above] {$b$};353 \draw[paperaccent,thick,domain=-.94:.94] plot (\x,{\x+1});